constify ether_sprintf and a few globals

[David Hill]

Hello -
This constify's ether_sprintf and some globals and adjusts the users
accordingly.

Feel free to commit if OK.

Index: dev/ic/if_wi_hostap.c
===
RCS file: /cvs/src/sys/dev/ic/if_wi_hostap.c,v
retrieving revision 1.52
diff -u -p -r1.52 if_wi_hostap.c
--- dev/ic/if_wi_hostap.c   19 Feb 2018 08:59:52 -  1.52
+++ dev/ic/if_wi_hostap.c   14 Dec 2022 15:32:35 -
@@ -76,13 +76,13 @@ struct wihap_sta_info *wihap_sta_find(st
 int wihap_sta_is_assoc(struct wihap_info *whi, u_int8_t addr[]);
 void wihap_auth_req(struct wi_softc *sc, struct wi_frame *rxfrm,
 caddr_t pkt, int len);
-void wihap_sta_deauth(struct wi_softc *sc, u_int8_t sta_addr[],
+void wihap_sta_deauth(struct wi_softc *sc, const u_int8_t sta_addr[],
 u_int16_t reason);
 void wihap_deauth_req(struct wi_softc *sc, struct wi_frame *rxfrm,
 caddr_t pkt, int len);
 void wihap_assoc_req(struct wi_softc *sc, struct wi_frame *rxfrm,
 caddr_t pkt, int len);
-void wihap_sta_disassoc(struct wi_softc *sc, u_int8_t sta_addr[],
+void wihap_sta_disassoc(struct wi_softc *sc, const u_int8_t sta_addr[],
 u_int16_t reason);
 void wihap_disassoc_req(struct wi_softc *sc, struct wi_frame *rxfrm,
 caddr_t pkt, int len);
@@ -204,7 +204,7 @@ wihap_init(struct wi_softc *sc)
  * Send a disassociation frame to a specified station.
  */
 void
-wihap_sta_disassoc(struct wi_softc *sc, u_int8_t sta_addr[], u_int16_t 
reason)
+wihap_sta_disassoc(struct wi_softc *sc, const u_int8_t sta_addr[], 
u_int16_t reason)

 {
struct wi_80211_hdr *resp_hdr;
caddr_t pkt;
@@ -233,7 +233,7 @@ wihap_sta_disassoc(struct wi_softc *sc,
  * Send a deauthentication message to a specified station.
  */
 void
-wihap_sta_deauth(struct wi_softc *sc, u_int8_t sta_addr[], u_int16_t 
reason)
+wihap_sta_deauth(struct wi_softc *sc, const u_int8_t sta_addr[], 
u_int16_t reason)

 {
struct wi_80211_hdr *resp_hdr;
caddr_t pkt;
Index: dev/pci/if_bwfm_pci.c
===
RCS file: /cvs/src/sys/dev/pci/if_bwfm_pci.c,v
retrieving revision 1.74
diff -u -p -r1.74 if_bwfm_pci.c
--- dev/pci/if_bwfm_pci.c   10 Nov 2022 12:14:05 -  1.74
+++ dev/pci/if_bwfm_pci.c   14 Dec 2022 15:32:35 -
@@ -1929,7 +1929,7 @@ bwfm_pci_flowring_lookup(struct bwfm_pci
 {
struct ieee80211com *ic = >sc_sc.sc_ic;
 #ifndef IEEE80211_STA_ONLY
-   uint8_t *da = mtod(m, uint8_t *);
+   const uint8_t *da = mtod(m, uint8_t *);
 #endif
int flowid, prio, fifo;
int i, found;
@@ -1987,7 +1987,7 @@ bwfm_pci_flowring_create(struct bwfm_pci
struct ieee80211com *ic = >sc_sc.sc_ic;
struct bwfm_cmd_flowring_create cmd;
 #ifndef IEEE80211_STA_ONLY
-   uint8_t *da = mtod(m, uint8_t *);
+   const uint8_t *da = mtod(m, uint8_t *);
 #endif
struct bwfm_pci_msgring *ring;
int flowid, prio, fifo;
Index: dev/pci/if_ixl.c
===
RCS file: /cvs/src/sys/dev/pci/if_ixl.c,v
retrieving revision 1.84
diff -u -p -r1.84 if_ixl.c
--- dev/pci/if_ixl.c5 Aug 2022 13:57:16 -   1.84
+++ dev/pci/if_ixl.c14 Dec 2022 15:32:35 -
@@ -1346,7 +1346,7 @@ static intixl_set_vsi(struct ixl_softc
 static int ixl_get_link_status(struct ixl_softc *);
 static int ixl_set_link_status(struct ixl_softc *,
const struct ixl_aq_desc *);
-static int ixl_add_macvlan(struct ixl_softc *, uint8_t *, uint16_t,
+static int ixl_add_macvlan(struct ixl_softc *, const uint8_t *, uint16_t,
uint16_t);
 static int ixl_remove_macvlan(struct ixl_softc *, uint8_t *, uint16_t,
uint16_t);
@@ -4547,7 +4547,7 @@ ixl_restart_an(struct ixl_softc *sc)
 }

 static int
-ixl_add_macvlan(struct ixl_softc *sc, uint8_t *macaddr, uint16_t vlan, 
uint16_t flags)
+ixl_add_macvlan(struct ixl_softc *sc, const uint8_t *macaddr, uint16_t 
vlan, uint16_t flags)

 {
struct ixl_aq_desc iaq;
struct ixl_aq_add_macvlan *param;
Index: dev/pci/if_mcx.c
===
RCS file: /cvs/src/sys/dev/pci/if_mcx.c,v
retrieving revision 1.106
diff -u -p -r1.106 if_mcx.c
--- dev/pci/if_mcx.c22 Nov 2022 06:48:32 -  1.106
+++ dev/pci/if_mcx.c14 Dec 2022 15:32:36 -
@@ -2564,7 +2564,7 @@ static intmcx_create_flow_group(struct
int, int, struct mcx_flow_match *);
 static int mcx_destroy_flow_group(struct mcx_softc *, int);
 static int mcx_set_flow_table_entry_mac(struct mcx_softc *, int, int,
-   uint8_t *, uint32_t);
+   const uint8_t *, uint32_t);
 static int mcx_set_flow_table_entry_proto(struct mcx_softc *, int, int,
int, int, uint32_t);
 static int mcx_delete_flow_table_entry(struct 

Re: splassert on boot

[David Hill]

On 11/23/22 15:18, Vitaliy Makkoveev wrote:

On Wed, Nov 23, 2022 at 02:59:05PM -0500, David Hill wrote:

Hello -

I am seeing splasserts on boot (before kern.splassert=2 can be set) with
-current.



spdmem0 at iic0 addr 0x50: 8GB DDR3 SDRAM PC3-12800 SO-DIMM
isa0 at pcib0
isadma0 at isa0
vga0 at isa0 port 0x3b0/48 iomem 0xa/131072
wsdisplay at vga0 not configured
pcppi0 at isa0 port 0x61
spkr0 at pcppi0
vmm0 at mainbus0: VMX/EPT (using slow L1TF mitigation)
splassert: pfi_attach_ifgroup: want 2 have 0
splassert: pfi_group_addmember: want 2 have 0
splassert: pfi_attach_ifgroup: want 2 have 0
splassert: pfi_attach_ifgroup: want 2 have 0
splassert: pfi_group_addmember: want 2 have 0


- David


The netlock assertion within PF_LOCK() looks wrong. The netlock should
be taken first, but only if both locks taken.

Index: sys/net/pfvar_priv.h
===
RCS file: /cvs/src/sys/net/pfvar_priv.h,v
retrieving revision 1.21
diff -u -p -r1.21 pfvar_priv.h
--- sys/net/pfvar_priv.h11 Nov 2022 17:12:30 -  1.21
+++ sys/net/pfvar_priv.h23 Nov 2022 20:14:13 -
@@ -278,7 +278,6 @@ extern struct rwlockpf_lock;
  extern struct rwlock  pf_state_lock;
  
  #define PF_LOCK()		do {			\

-   NET_ASSERT_LOCKED();\
rw_enter_write(_lock);   \
} while (0)
  


with traceback

vmm0 at mainbus0: VMX/EPT (using slow L1TF mitigation)
splassert: pfi_attach_ifgroup: want 2 have 0
Starting stack trace...
pfi_attach_ifgroup(800ecc80,800ecc80,81f25c9b,800ecca8,8188cf82,82737e30) 
at pfi_attach_ifgroup+0x42
if_addgroup(8015,81f25c9b,f9c039b9154dadd0,0,8015,822d06e0) 
at if_addgroup+0x142
enc_clone_create(822d06e0,0,28d493f4921b481d,2,822e5aa0,1f) 
at enc_clone_create+0xd2
encattach(1,1,8173c274,82737f00,822d06e0,0) at 
encattach+0x14

main(0,0,2,0,abd41b90,1) at main+0x409
end trace frame: 0x0, count: 252
End of stack trace.
splassert: pfi_group_addmember: want 2 have 0
Starting stack trace...
pfi_group_addmember(81f25c9b,81f25c9b,81f25c9b,8014a228,8188d442,82737e30) 
at pfi_group_addmember+0x42
if_addgroup(8015,81f25c9b,f9c039b9154dadd0,0,8015,822d06e0) 
at if_addgroup+0x1e6
enc_clone_create(822d06e0,0,28d493f4921b481d,2,822e5aa0,1f) 
at enc_clone_create+0xd2
encattach(1,1,8173c274,82737f00,822d06e0,0) at 
encattach+0x14

main(0,0,2,0,abd41b90,1) at main+0x409
end trace frame: 0x0, count: 252
End of stack trace.
splassert: pfi_attach_ifgroup: want 2 have 0
Starting stack trace...
pfi_attach_ifgroup(80102e80,80102e80,80102e80,80102ea8,8188cf82,82737e90) 
at pfi_attach_ifgroup+0x42
if_creategroup(81f501b2,81f501b2,ea09b7f381ae3cd3,2,822e5b50,822a6790) 
at if_creategroup+0x6f
carpattach(1,1,e1e0b7d5818b2660,822a6790,811da510,82737f00) 
at carpattach+0x20

main(0,0,2,0,abd41b90,1) at main+0x409
end trace frame: 0x0, count: 253
End of stack trace.
splassert: pfi_attach_ifgroup: want 2 have 0
Starting stack trace...
pfi_attach_ifgroup(800f2f00,800f2f00,81f77eea,800f2f28,8188cf82,82737e70) 
at pfi_attach_ifgroup+0x42
if_addgroup(80150800,81f77eea,f9c039b91557e57e,2,80150800,1f) 
at if_addgroup+0x142
loopattach(1,1,822e5b90,822db440,8169888c,82737f00) 
at loopattach+0xbc

main(0,0,2,0,abd41b90,1) at main+0x409
end trace frame: 0x0, count: 253
End of stack trace.
splassert: pfi_group_addmember: want 2 have 0
Starting stack trace...
pfi_group_addmember(81f77eea,81f77eea,81f77eea,8014a2e8,8188d442,82737e70) 
at pfi_group_addmember+0x42
if_addgroup(80150800,81f77eea,f9c039b91557e57e,2,80150800,1f) 
at if_addgroup+0x1e6
loopattach(1,1,822e5b90,822db440,8169888c,82737f00) 
at loopattach+0xbc

main(0,0,2,0,abd41b90,1) at main+0x409
end trace frame: 0x0, count: 253
End of stack trace.

splassert on boot

[David Hill]

Hello -

I am seeing splasserts on boot (before kern.splassert=2 can be set) with 
-current.




spdmem0 at iic0 addr 0x50: 8GB DDR3 SDRAM PC3-12800 SO-DIMM
isa0 at pcib0
isadma0 at isa0
vga0 at isa0 port 0x3b0/48 iomem 0xa/131072
wsdisplay at vga0 not configured
pcppi0 at isa0 port 0x61
spkr0 at pcppi0
vmm0 at mainbus0: VMX/EPT (using slow L1TF mitigation)
splassert: pfi_attach_ifgroup: want 2 have 0
splassert: pfi_group_addmember: want 2 have 0
splassert: pfi_attach_ifgroup: want 2 have 0
splassert: pfi_attach_ifgroup: want 2 have 0
splassert: pfi_group_addmember: want 2 have 0


- David

Re: possible fix for Xorg 21.1.1 crashes

[David Hill]

On 11/16/21 18:21, Matthieu Herrb wrote:

On Tue, Nov 16, 2021 at 11:26:40PM +0100, Matthieu Herrb wrote:

Hi,

I think I found the bug that causes crashes in X for some people.

If X started crashing since you upgraded to the last snapshots, can
you try the patch below ?

get /usr/xenocara from CVS then

cd /usr/xenocara/xserver
patch -p0 -E < /this/patch
doas make -f Makefile.bsd-wrapper obj
doas make -f Makefile.bsd-wrapper build

And restart the X server, for example :
doas rcctl restart xenodm


Here's a more complete fix. Ust this one instead, sorry.



The first diff did not help, but this one has kept my X stable for 12 
hours so far...


- David

postgresql 12 - could not open shared memory segment

[David Hill]

Hello -

I am using postgresql 12 on OpenBSD 6.8.  After a while of running, 
postgresql starts throwing out errors:


2020-12-03 08:18:35.052 CST [60288] ERROR:  could not open shared memory 
segment "/PostgreSQL.1780529043": No such file or directory


2020-12-03 08:18:35.052 CST [40083] ERROR:  could not open shared memory 
segment "/PostgreSQL.1780529043": No such file or directory


2020-12-03 08:18:35.052 CST [53517] ERROR:  could not open shared memory 
segment "/PostgreSQL.1780529043": No such file or directory


2020-12-03 08:18:35.052 CST [53517] CONTEXT:  parallel worker

2020-12-03 08:18:35.054 CST [47581] LOG:  background worker "parallel 
worker" (PID 40083) exited with exit code 1


2020-12-03 08:18:35.055 CST [47581] LOG:  background worker "parallel 
worker" (PID 60288) exited with exit code 1


2020-12-03 08:24:19.583 CST [53517] WARNING:  could not remove shared 
memory segment "/PostgreSQL.1222503421": No such file or directory


2020-12-03 08:24:19.584 CST [31517] WARNING:  could not remove shared 
memory segment "/PostgreSQL.19516083": No such file or directory


2020-12-03 08:24:19.584 CST [53517] WARNING:  could not remove shared 
memory segment "/PostgreSQL.1780529043": No such file or directory



in postgresql.conf, dynamic_shared_memory_type is set to posix.

/etc/sysctl.conf
kern.seminfo.semmni=256
kern.seminfo.semmns=2048

I am wondering there is something I need to adjust... or if anyone has 
run into this before?  Does anyone have any pointers on what to look for?


Thanks
- David

Re: monotonic clock going backwards?

[David Hill]

On 6/6/19 8:49 AM, Mark Kettenis wrote:

From: David Hill 
Date: Thu, 6 Jun 2019 08:22:06 -0400

On 6/6/19 6:56 AM, Scott Cheloha wrote:

On Wed, Jun 05, 2019 at 08:33:56PM -0400, David Hill wrote:

Hello -

I noticed some negative roundtrip times when running traceroute, which uses
the monotonic clock to calculate the RTT.

If I run the following code, it eventually bombs.?? It bombs quick if I
launch Firefox.

timespeccmp failed
tp1 s:103780 n:63101042
tp2 s:103779 n:761117849

please share your dmesg and the output of

sysctl kern.clockrate kern.timecounter

Could you also share the output when you run the following code?
With and without firefox running?  Curious about how frequently
this happens.

#include 

#include 
#include 
#include 

int
main(void)
{
struct timespec t0, t1, timeout;

timeout.tv_sec = 0;
timeout.tv_nsec = 10;

for (;;) {
clock_gettime(CLOCK_MONOTONIC, );
if (nanosleep(, NULL) == -1)
err(1, "nanosleep");
clock_gettime(CLOCK_MONOTONIC, );
if (timespeccmp(, , >)) {
printf("%lld.%09ld precedes %lld.%09ld\n",
t1.tv_sec, t1.tv_nsec, t0.tv_sec, t0.tv_nsec);
}
}

return 0;
}


$ sysctl kern.clockrate kern.timecounter
kern.clockrate=tick = 1, tickadj = 40, hz = 100, profhz = 100,
stathz = 100
kern.timecounter.tick=1
kern.timecounter.timestepwarnings=0
kern.timecounter.hardware=tsc
kern.timecounter.choice=i8254(0) acpihpet0(1000) tsc(2000) acpitimer0(1000)

146259.148543164 precedes 146259.450584680
146259.488592813 precedes 146259.790556528
146259.808566236 precedes 146260.110528202
146261.148476141 precedes 146261.450398822
146262.888279505 precedes 146263.190240715
146263.887431808 precedes 146264.189435801
146274.487159451 precedes 146274.789148187
146278.186820816 precedes 146278.488803756
146279.546680181 precedes 146279.848674138
146284.326305557 precedes 146284.628224696
146285.426154861 precedes 146285.728066901
146287.885921929 precedes 146288.187826688
146289.985763711 precedes 146290.287731233
146290.985659041 precedes 146291.287640302
146293.985348615 precedes 146294.287327519
146295.465186742 precedes 146295.767218516
146295.885156426 precedes 146296.187140717
146298.544907956 precedes 146298.846882759
146298.644894308 precedes 146298.946889142

Does this also happen with

kern.timecounter.hardware=acpihpet

?


Ran the code with kern.timecounter.hardware=acpihpet0 for 10 minutes 
without any output.


With the code still running, I switched to tsc and starting seeing output:

151944.345840447 precedes 151944.647753895
151944.525774690 precedes 151944.827681380
151946.405553296 precedes 151946.707533012

The output stops as soon as I switch back to acpihpet0.

I will add that to my /etc/sysctl.conf.

Thank you.





OpenBSD 6.5-current (GENERIC.MP) #27: Tue Jun  4 15:36:14 EDT 2019
dh...@liver.mindcry.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP
real mem = 8418242560 (8028MB)
avail mem = 8152993792 (7775MB)
mpath0 at root
scsibus0 at mpath0: 256 targets
mainbus0 at root
bios0 at mainbus0: SMBIOS rev. 2.8 @ 0x9a565000 (49 entries)
bios0: vendor American Megatrends Inc. version "X420UA.204" date 11/16/2018
bios0: ASUSTeK COMPUTER INC. VivoBook_ASUSLaptop X420UA
acpi0 at bios0: rev 2
acpi0: sleep states S0 S3 S4 S5
acpi0: tables DSDT FACP APIC FPDT FIDT ECDT MCFG SSDT SSDT SSDT MSDM
SSDT HPET SSDT UEFI SSDT LPIT SSDT SSDT SSDT SSDT DBGP DBG2 DMAR NHLT
BGRT TPM2 WSMT
acpi0: wakeup devices PXSX(S4) RP06(S4) GLAN(S4) XHC_(S3) XDCI(S4) HDAS(S4)
acpitimer0 at acpi0: 3579545 Hz, 24 bits
acpimadt0 at acpi0 addr 0xfee0: PC-AT compat
cpu0 at mainbus0: apid 0 (boot processor)
cpu0: Intel(R) Core(TM) i5-8250U CPU @ 1.60GHz, 3393.27 MHz, 06-8e-0a
cpu0:
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE,SSE3,PCLMUL,DTES64,MWAIT,DS-CPL,VMX,EST,TM2,S
SSE3,SDBG,FMA3,CX16,xTPR,PDCM,PCID,SSE4.1,SSE4.2,x2APIC,MOVBE,POPCNT,DEADLINE,AES,XSAVE,AVX,F16C,RDRAND,NXE,PAGE1GB,RDTSCP,LONG,LAHF,ABM,3DNOWP,PERF,ITSC,FSGSBASE,SGX,BMI
1,AVX2,SMEP,BMI2,ERMS,INVPCID,MPX,RDSEED,ADX,SMAP,CLFLUSHOPT,PT,MD_CLEAR,TSXFA,IBRS,IBPB,STIBP,L1DF,SSBD,SENSOR,ARAT,XSAVEOPT,XSAVEC,XGETBV1,XSAVES,MELTDOWN
cpu0: 256KB 64b/line 8-way L2 cache
cpu0: smt 0, core 0, package 0
mtrr: Pentium Pro MTRR support, 10 var ranges, 88 fixed ranges
cpu0: apic clock running at 23MHz
cpu0: mwait min=64, max=64, C-substates=0.2.1.2.4.1.1.1, IBE
cpu1 at mainbus0: apid 2 (application processor)
cpu1: Intel(R) Core(TM) i5-8250U CPU @ 1.60GHz, 3392.09 MHz, 06-8e-0a
cpu1:
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE,SSE3,PCLMUL,DTES64,MWAIT,DS-CPL,VMX,EST,TM2,S
SSE3,SDBG,FMA3,CX16,xTPR,PDCM,PCID,SSE4.1,SSE4.2,x2APIC,MOVBE,POPCNT,DEADLINE,AES,XSAVE,AVX,F1

Re: monotonic clock going backwards?

[David Hill]

On 6/6/19 6:56 AM, Scott Cheloha wrote:

On Wed, Jun 05, 2019 at 08:33:56PM -0400, David Hill wrote:

Hello -

I noticed some negative roundtrip times when running traceroute, which uses
the monotonic clock to calculate the RTT.

If I run the following code, it eventually bombs.?? It bombs quick if I
launch Firefox.

timespeccmp failed
tp1 s:103780 n:63101042
tp2 s:103779 n:761117849

please share your dmesg and the output of

sysctl kern.clockrate kern.timecounter

Could you also share the output when you run the following code?
With and without firefox running?  Curious about how frequently
this happens.

#include 

#include 
#include 
#include 

int
main(void)
{
struct timespec t0, t1, timeout;

timeout.tv_sec = 0;
timeout.tv_nsec = 10;

for (;;) {
clock_gettime(CLOCK_MONOTONIC, );
if (nanosleep(, NULL) == -1)
err(1, "nanosleep");
clock_gettime(CLOCK_MONOTONIC, );
if (timespeccmp(, , >)) {
printf("%lld.%09ld precedes %lld.%09ld\n",
t1.tv_sec, t1.tv_nsec, t0.tv_sec, t0.tv_nsec);
}
}

return 0;
}



$ sysctl kern.clockrate kern.timecounter
kern.clockrate=tick = 1, tickadj = 40, hz = 100, profhz = 100, 
stathz = 100

kern.timecounter.tick=1
kern.timecounter.timestepwarnings=0
kern.timecounter.hardware=tsc
kern.timecounter.choice=i8254(0) acpihpet0(1000) tsc(2000) acpitimer0(1000)

146259.148543164 precedes 146259.450584680
146259.488592813 precedes 146259.790556528
146259.808566236 precedes 146260.110528202
146261.148476141 precedes 146261.450398822
146262.888279505 precedes 146263.190240715
146263.887431808 precedes 146264.189435801
146274.487159451 precedes 146274.789148187
146278.186820816 precedes 146278.488803756
146279.546680181 precedes 146279.848674138
146284.326305557 precedes 146284.628224696
146285.426154861 precedes 146285.728066901
146287.885921929 precedes 146288.187826688
146289.985763711 precedes 146290.287731233
146290.985659041 precedes 146291.287640302
146293.985348615 precedes 146294.287327519
146295.465186742 precedes 146295.767218516
146295.885156426 precedes 146296.187140717
146298.544907956 precedes 146298.846882759
146298.644894308 precedes 146298.946889142

OpenBSD 6.5-current (GENERIC.MP) #27: Tue Jun  4 15:36:14 EDT 2019
dh...@liver.mindcry.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP
real mem = 8418242560 (8028MB)
avail mem = 8152993792 (7775MB)
mpath0 at root
scsibus0 at mpath0: 256 targets
mainbus0 at root
bios0 at mainbus0: SMBIOS rev. 2.8 @ 0x9a565000 (49 entries)
bios0: vendor American Megatrends Inc. version "X420UA.204" date 11/16/2018
bios0: ASUSTeK COMPUTER INC. VivoBook_ASUSLaptop X420UA
acpi0 at bios0: rev 2
acpi0: sleep states S0 S3 S4 S5
acpi0: tables DSDT FACP APIC FPDT FIDT ECDT MCFG SSDT SSDT SSDT MSDM 
SSDT HPET SSDT UEFI SSDT LPIT SSDT SSDT SSDT SSDT DBGP DBG2 DMAR NHLT 
BGRT TPM2 WSMT

acpi0: wakeup devices PXSX(S4) RP06(S4) GLAN(S4) XHC_(S3) XDCI(S4) HDAS(S4)
acpitimer0 at acpi0: 3579545 Hz, 24 bits
acpimadt0 at acpi0 addr 0xfee0: PC-AT compat
cpu0 at mainbus0: apid 0 (boot processor)
cpu0: Intel(R) Core(TM) i5-8250U CPU @ 1.60GHz, 3393.27 MHz, 06-8e-0a
cpu0: 
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE,SSE3,PCLMUL,DTES64,MWAIT,DS-CPL,VMX,EST,TM2,S

SSE3,SDBG,FMA3,CX16,xTPR,PDCM,PCID,SSE4.1,SSE4.2,x2APIC,MOVBE,POPCNT,DEADLINE,AES,XSAVE,AVX,F16C,RDRAND,NXE,PAGE1GB,RDTSCP,LONG,LAHF,ABM,3DNOWP,PERF,ITSC,FSGSBASE,SGX,BMI
1,AVX2,SMEP,BMI2,ERMS,INVPCID,MPX,RDSEED,ADX,SMAP,CLFLUSHOPT,PT,MD_CLEAR,TSXFA,IBRS,IBPB,STIBP,L1DF,SSBD,SENSOR,ARAT,XSAVEOPT,XSAVEC,XGETBV1,XSAVES,MELTDOWN
cpu0: 256KB 64b/line 8-way L2 cache
cpu0: smt 0, core 0, package 0
mtrr: Pentium Pro MTRR support, 10 var ranges, 88 fixed ranges
cpu0: apic clock running at 23MHz
cpu0: mwait min=64, max=64, C-substates=0.2.1.2.4.1.1.1, IBE
cpu1 at mainbus0: apid 2 (application processor)
cpu1: Intel(R) Core(TM) i5-8250U CPU @ 1.60GHz, 3392.09 MHz, 06-8e-0a
cpu1: 
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE,SSE3,PCLMUL,DTES64,MWAIT,DS-CPL,VMX,EST,TM2,S

SSE3,SDBG,FMA3,CX16,xTPR,PDCM,PCID,SSE4.1,SSE4.2,x2APIC,MOVBE,POPCNT,DEADLINE,AES,XSAVE,AVX,F16C,RDRAND,NXE,PAGE1GB,RDTSCP,LONG,LAHF,ABM,3DNOWP,PERF,ITSC,FSGSBASE,SGX,BMI
1,AVX2,SMEP,BMI2,ERMS,INVPCID,MPX,RDSEED,ADX,SMAP,CLFLUSHOPT,PT,MD_CLEAR,TSXFA,IBRS,IBPB,STIBP,L1DF,SSBD,SENSOR,ARAT,XSAVEOPT,XSAVEC,XGETBV1,XSAVES,MELTDOWN
cpu1: 256KB 64b/line 8-way L2 cache
cpu1: smt 0, core 1, package 0
cpu2 at mainbus0: apid 4 (application processor)
cpu2: Intel(R) Core(TM) i5-8250U CPU @ 1.60GHz, 3392.09 MHz, 06-8e-0a
cpu2: 
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE,SSE3,PCLMUL,DTES64

monotonic clock going backwards?

[David Hill]

Hello -

I noticed some negative roundtrip times when running traceroute, which 
uses the monotonic clock to calculate the RTT.


If I run the following code, it eventually bombs.  It bombs quick if I 
launch Firefox.


timespeccmp failed
tp1 s:103780 n:63101042
tp2 s:103779 n:761117849





#include 
#include 
#include 
#include 

int
main(void)
{
    int r;
    struct timespec tp1, tp2, tout;

    tout.tv_sec = 0;
    tout.tv_nsec = 10;

    for (;;) {
    r = clock_gettime(CLOCK_MONOTONIC, );
    if (r == -1) {
    perror("clock_gettime");
    exit(-1);
    }

    nanosleep(, NULL);

    r = clock_gettime(CLOCK_MONOTONIC, );
    if (r == -1) {
    perror("clock_gettime");
    exit(-1);
    }

    // tp1 should never be larger than tp2
    r = timespeccmp(, , >);
    if (r == 1) {
    printf("timespeccmp failed\n");
    printf("tp1 s:%lld n:%ld\n", tp1.tv_sec, 
tp1.tv_nsec);
    printf("tp2 s:%lld n:%ld\n", tp2.tv_sec, 
tp2.tv_nsec);

    exit(-1);
    }
    }

    return 0;
}

Re: iked: fix retransmission bug after simultaneous rekeying

[David Hill]

On Thu, Apr 11, 2019 at 04:32:28PM +0200, Tobias Heider wrote:
> Hi,
> 
> this patch fixes a bug that appears after simultaneous
> rekeying of the ikesa. Currently the initiator does not set it's
> IKED_REQ_INFORMATIONAL flag when sending the delete request and thus rejects
> the response to the DELETE INFORMATIONAL request in ikev2_recv.
> The request is not removed from the initiators request queue and is
> retransmitted over and over.
> 
> The bug can best be reproduced by configuring two iked peers with a short
> ikelifetime (e.g. 10) and looking for INFORMATIONAL receives on the
> responder side.
> 
> 
> Index: sbin/iked//ikev2.c
> ===
> RCS file: /mount/openbsd/cvs/src/sbin/iked/ikev2.c,v
> retrieving revision 1.168
> diff -u -p -u -r1.168 ikev2.c
> --- sbin/iked//ikev2.c27 Feb 2019 06:33:56 -  1.168
> +++ sbin/iked//ikev2.c11 Apr 2019 14:09:35 -
> @@ -3539,6 +3539,9 @@ ikev2_ikesa_delete(struct iked *env, str
>   struct ikev2_delete *del;
>  
>   if (initiator) {
> + /* XXX: Can not have simultaneous INFORMATIONAL exchanges */
> + if (sa->sa_stateflags & IKED_REQ_INF)
> + goto done;
>   /* Send PAYLOAD_DELETE */
>   if ((buf = ibuf_static()) == NULL)
>   goto done;
> @@ -3550,6 +3553,7 @@ ikev2_ikesa_delete(struct iked *env, str
>   if (ikev2_send_ike_e(env, sa, buf, IKEV2_PAYLOAD_DELETE,
>   IKEV2_EXCHANGE_INFORMATIONAL, 0) == -1)
>   goto done;
> + sa->sa_stateflags |= IKED_REQ_INF;
>   log_debug("%s: sent delete, closing SA", __func__);
>  done:
>   ibuf_release(buf);
>

This makes my iked connections much more stable.

- David

diff - set ipv6 tclass

[David Hill]

Hello -

ftp, ntpd, and telnet set IPTOS_LOWDELAY/IPTOS_THROUGHPUT for IPv4
sockets.  Should we do it for IPv6 as well?  I used
IPTOS_DCSP_AF21/IPTOS_DSCP_CS1 to match the values ssh uses by default.

Thoughts?

Index: usr.bin/ftp/ftp.c
===
RCS file: /cvs/src/usr.bin/ftp/ftp.c,v
retrieving revision 1.100
diff -u -p -r1.100 ftp.c
--- usr.bin/ftp/ftp.c   22 Aug 2016 16:27:00 -  1.100
+++ usr.bin/ftp/ftp.c   30 Oct 2018 19:31:34 -
@@ -257,10 +257,15 @@ hookup(char *host, char *port)
code = -1;
goto bad;
}
-   if (hisctladdr.sa.sa_family == AF_INET) {
+   switch (hisctladdr.sa.sa_family) {
+   case AF_INET:
tos = IPTOS_LOWDELAY;
-   if (setsockopt(s, IPPROTO_IP, IP_TOS, (char *), 
sizeof(int)) < 0)
-   warn("setsockopt TOS (ignored)");
+   if (setsockopt(s, IPPROTO_IP, IP_TOS, , sizeof(tos)) < 0)
+   warn("setsockopt IPTOS_LOWDELAY");
+   case AF_INET6:
+   tos = IPTOS_DSCP_AF21;
+   if (setsockopt(s, IPPROTO_IPV6, IPV6_TCLASS, , sizeof(tos)) 
< 0)
+   warn("setsockopt IPTOS_DSCP_AF21");
}
cin = fdopen(s, "r");
cout = fdopen(s, "w");
@@ -1524,11 +1529,17 @@ reinit:
warn("connect");
goto bad;
}
-   if (data_addr.sa.sa_family == AF_INET) {
+   switch (data_addr.sa.sa_family) {
+   case AF_INET:
on = IPTOS_THROUGHPUT;
-   if (setsockopt(data, IPPROTO_IP, IP_TOS, (char *),
-  sizeof(int)) < 0)
-   warn("setsockopt TOS (ignored)");
+   if (setsockopt(data, IPPROTO_IP, IP_TOS, ,
+   sizeof(on)) < 0)
+   warn("setsockopt IPTOS_THROUGHPUT");
+   case AF_INET6:
+   on = IPTOS_DSCP_CS1;
+   if (setsockopt(data, IPPROTO_IPV6, IPV6_TCLASS, ,
+   sizeof(on)) < 0)
+   warn("setsockopt IPTOS_DSCP_CS1");
}
return (0);
}
@@ -1682,7 +1693,7 @@ dataconn(const char *lmode)
 {
union sockaddr_union from;
socklen_t fromlen = myctladdr.sa.sa_len;
-   int s;
+   int s, tos;
 
if (passivemode)
return (fdopen(data, lmode));
@@ -1695,12 +1706,16 @@ dataconn(const char *lmode)
}
(void)close(data);
data = s;
-   if (from.sa.sa_family == AF_INET) {
-   int tos = IPTOS_THROUGHPUT;
-   if (setsockopt(s, IPPROTO_IP, IP_TOS, (char *),
-   sizeof(int)) < 0) {
-   warn("setsockopt TOS (ignored)");
-   }
+   switch (from.sa.sa_family) {
+   case AF_INET:
+   tos = IPTOS_THROUGHPUT;
+   if (setsockopt(s, IPPROTO_IP, IP_TOS, , sizeof(tos)) < 0)
+   warn("setsockopt IPTOS_THROUGHPUT");
+   case AF_INET6:
+   tos = IPTOS_DSCP_CS1;
+   if (setsockopt(s, IPPROTO_IPV6, IPV6_TCLASS, ,
+   sizeof(tos)) < 0)
+   warn("setsockopt IPTOS_DSCP_CS1");
}
return (fdopen(data, lmode));
 }
Index: usr.bin/telnet/commands.c
===
RCS file: /cvs/src/usr.bin/telnet/commands.c,v
retrieving revision 1.86
diff -u -p -r1.86 commands.c
--- usr.bin/telnet/commands.c   30 Sep 2018 14:35:32 -  1.86
+++ usr.bin/telnet/commands.c   30 Oct 2018 19:31:35 -
@@ -1742,8 +1742,8 @@ tn(int argc, char *argv[])
 {
 struct addrinfo hints, *res, *res0;
 char *cmd, *hostp = 0, *portp = 0, *user = 0, *aliasp = 0;
-int error, retry;
-const int niflags = NI_NUMERICHOST, tos = IPTOS_LOWDELAY;
+int error, retry, tos;
+const int niflags = NI_NUMERICHOST;
 
 if (connected) {
printf("?Already connected to %s\r\n", hostname);
@@ -1870,11 +1870,13 @@ tn(int argc, char *argv[])
 
switch (res->ai_family) {
case AF_INET:
+   tos = IPTOS_LOWDELAY;
if (setsockopt(net, IPPROTO_IP, IP_TOS, , sizeof(tos)) < 0
&& errno != ENOPROTOOPT)
perror("telnet: setsockopt (IP_TOS) (ignored)");
break;
case AF_INET6:
+   tos = IPTOS_DSCP_AF21;
if (setsockopt(net, IPPROTO_IPV6, IPV6_TCLASS, ,
sizeof(tos)) < 0 && errno != ENOPROTOOPT)
perror("telnet: setsockopt (IPV6_TCLASS) (ignored)");
Index: usr.sbin/ntpd/client.c
===
RCS file: /cvs/src/usr.sbin/ntpd/client.c,v
retrieving revision 1.105
diff 

diff: split ctloutput into getopt/setopt

[David Hill]

Hello -

This diff splits the ctloutput functions into getopt/setopt, which could
offer more fine-grained locking.  It also removes some indentation and
imo is easier to read.

Thoughts?

Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.279
diff -u -p -r1.279 rtsock.c
--- net/rtsock.c10 Jul 2018 20:28:34 -  1.279
+++ net/rtsock.c18 Oct 2018 15:26:28 -
@@ -110,6 +110,8 @@ voidrcb_unref(void *, void *);
 introute_output(struct mbuf *, struct socket *, struct sockaddr *,
struct mbuf *);
 introute_ctloutput(int, struct socket *, int, int, struct mbuf *);
+introute_getopt(struct socket *, int, int, struct mbuf *);
+introute_setopt(struct socket *, int, int, struct mbuf *);
 introute_usrreq(struct socket *, int, struct mbuf *, struct mbuf *,
struct mbuf *, struct proc *);
 void   route_input(struct mbuf *m0, struct socket *, sa_family_t);
@@ -358,69 +360,98 @@ int
 route_ctloutput(int op, struct socket *so, int level, int optname,
 struct mbuf *m)
 {
+   int error;
+
+   switch (op) {
+   case PRCO_SETOPT:
+   error = route_setopt(so, level, optname, m);
+   break;
+   case PRCO_GETOPT:
+   error = route_getopt(so, level, optname, m);
+   break;
+   default:
+   error = EINVAL;
+   break;
+   }
+
+   return error;
+}
+
+int
+route_setopt(struct socket *so, int level, int optname, struct mbuf *m)
+{
struct rtpcb *rop = sotortpcb(so);
int error = 0;
unsigned int tid, prio;
 
if (level != AF_ROUTE)
-   return (EINVAL);
+   return EINVAL;
 
-   switch (op) {
-   case PRCO_SETOPT:
-   switch (optname) {
-   case ROUTE_MSGFILTER:
-   if (m == NULL || m->m_len != sizeof(unsigned int))
-   error = EINVAL;
-   else
-   rop->rop_msgfilter = *mtod(m, unsigned int *);
-   break;
-   case ROUTE_TABLEFILTER:
-   if (m == NULL || m->m_len != sizeof(unsigned int)) {
-   error = EINVAL;
-   break;
-   }
-   tid = *mtod(m, unsigned int *);
-   if (tid != RTABLE_ANY && !rtable_exists(tid))
-   error = ENOENT;
-   else
-   rop->rop_rtableid = tid;
-   break;
-   case ROUTE_PRIOFILTER:
-   if (m == NULL || m->m_len != sizeof(unsigned int)) {
-   error = EINVAL;
-   break;
-   }
-   prio = *mtod(m, unsigned int *);
-   if (prio > RTP_MAX)
-   error = EINVAL;
-   else
-   rop->rop_priority = prio;
-   break;
-   default:
-   error = ENOPROTOOPT;
+   switch (optname) {
+   case ROUTE_MSGFILTER:
+   if (m == NULL || m->m_len != sizeof(unsigned int))
+   error = EINVAL;
+   else
+   rop->rop_msgfilter = *mtod(m, unsigned int *);
+   break;
+   case ROUTE_TABLEFILTER:
+   if (m == NULL || m->m_len != sizeof(unsigned int)) {
+   error = EINVAL;
break;
}
+   tid = *mtod(m, unsigned int *);
+   if (tid != RTABLE_ANY && !rtable_exists(tid))
+   error = ENOENT;
+   else
+   rop->rop_rtableid = tid;
break;
-   case PRCO_GETOPT:
-   switch (optname) {
-   case ROUTE_MSGFILTER:
-   m->m_len = sizeof(unsigned int);
-   *mtod(m, unsigned int *) = rop->rop_msgfilter;
-   break;
-   case ROUTE_TABLEFILTER:
-   m->m_len = sizeof(unsigned int);
-   *mtod(m, unsigned int *) = rop->rop_rtableid;
-   break;
-   case ROUTE_PRIOFILTER:
-   m->m_len = sizeof(unsigned int);
-   *mtod(m, unsigned int *) = rop->rop_priority;
-   break;
-   default:
-   error = ENOPROTOOPT;
+   case ROUTE_PRIOFILTER:
+   if (m == NULL || m->m_len != sizeof(unsigned int)) {
+   error = EINVAL;
break;
}
+   prio = *mtod(m, unsigned int *);
+   if (prio > RTP_MAX)
+   

re(4) watchdog timeouts - possible fix

[David Hill]

Hello -

If I start a certain app that sends a large amount of data to many
hosts, I get:

re0: watchdog timeout
re0: watchdog timeout

The diff below is what em(4) does and I am no longer able to cause
watchdog timeouts on my re(4).

Does this seem correct or more of a hack?

Index: dev/ic/re.c
===
RCS file: /cvs/src/sys/dev/ic/re.c,v
retrieving revision 1.202
diff -u -p -r1.202 re.c
--- dev/ic/re.c 19 Jun 2017 09:36:27 -  1.202
+++ dev/ic/re.c 31 Aug 2018 19:16:16 -
@@ -1817,6 +1817,9 @@ re_start(struct ifqueue *ifq)
bpf_mtap_ether(ifp->if_bpf, m, BPF_DIRECTION_OUT);
 #endif
 
+   /* Set timeout in case hardware has problems transmitting */
+   ifp->if_timer = 5;
+
KASSERT(used <= free);
free -= used;
 
@@ -1830,7 +1833,6 @@ re_start(struct ifqueue *ifq)
if (post == 0)
return;
 
-   ifp->if_timer = 5;
sc->rl_ldata.rl_txq_prodidx = idx;
ifq_serialize(ifq, >rl_start);
 }

more netinet bcopy -> mem{cpy|move}

[David Hill]

Hello -

Two bcopy->memmove conversions where src and dst are in the same memory
space.  Perhaps not overlapping, but just being safe.

Convert the bcopy's in tcp_respond to memcpy since the destination is a
new mbuf header.

OK?

Index: netinet/ip_ah.c
===
RCS file: /cvs/src/sys/netinet/ip_ah.c,v
retrieving revision 1.138
diff -u -p -r1.138 ip_ah.c
--- netinet/ip_ah.c 14 Mar 2018 22:38:46 -  1.138
+++ netinet/ip_ah.c 9 Apr 2018 17:48:20 -
@@ -899,8 +899,8 @@ ah_input_cb(struct cryptop *crp)
 * mbuf...do an overlapping copy of the
 * remainder of the mbuf over the ESP header.
 */
-   bcopy(mtod(m1, u_char *) + roff + rplen +
-   ahx->authsize, mtod(m1, u_char *) + roff,
+   memmove(mtod(m1, u_char *) + roff,
+   mtod(m1, u_char *) + roff + rplen + ahx->authsize,
m1->m_len - (roff + rplen + ahx->authsize));
m1->m_len -= rplen + ahx->authsize;
m->m_pkthdr.len -= rplen + ahx->authsize;
Index: netinet/tcp_subr.c
===
RCS file: /cvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.170
diff -u -p -r1.170 tcp_subr.c
--- netinet/tcp_subr.c  2 Apr 2018 14:19:17 -   1.170
+++ netinet/tcp_subr.c  9 Apr 2018 17:48:20 -
@@ -328,11 +328,11 @@ tcp_respond(struct tcpcb *tp, caddr_t te
th = (struct tcphdr *)(ip6 + 1);
tlen = sizeof(*ip6) + sizeof(*th);
if (th0) {
-   bcopy(template, ip6, sizeof(*ip6));
-   bcopy(th0, th, sizeof(*th));
+   memcpy(ip6, template, sizeof(*ip6));
+   memcpy(th, th0, sizeof(*th));
xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
} else {
-   bcopy(template, ip6, tlen);
+   memcpy(ip6, template, tlen);
}
break;
 #endif /* INET6 */
@@ -341,11 +341,11 @@ tcp_respond(struct tcpcb *tp, caddr_t te
th = (struct tcphdr *)(ip + 1);
tlen = sizeof(*ip) + sizeof(*th);
if (th0) {
-   bcopy(template, ip, sizeof(*ip));
-   bcopy(th0, th, sizeof(*th));
+   memcpy(ip, template, sizeof(*ip));
+   memcpy(th, th0, sizeof(*th));
xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, u_int32_t);
} else {
-   bcopy(template, ip, tlen);
+   memcpy(ip, template, tlen);
}
break;
}
Index: netinet/udp_usrreq.c
===
RCS file: /cvs/src/sys/netinet/udp_usrreq.c,v
retrieving revision 1.246
diff -u -p -r1.246 udp_usrreq.c
--- netinet/udp_usrreq.c6 Apr 2018 10:59:11 -   1.246
+++ netinet/udp_usrreq.c9 Apr 2018 17:48:20 -
@@ -295,8 +295,8 @@ udp_input(struct mbuf **mp, int *offp, i
}
 
/* remove the UDP header */
-   bcopy(mtod(m, u_char *),
-   mtod(m, u_char *) + sizeof(struct udphdr), iphlen);
+   memmove(mtod(m, u_char *) + sizeof(struct udphdr),
+   mtod(m, u_char *), iphlen);
m_adj(m, sizeof(struct udphdr));
skip -= sizeof(struct udphdr);
 

Re: ufs free()

[David Hill]

On Sat, Mar 31, 2018 at 11:15:35PM +0200, Christian Weisgerber wrote:
> David Hill:
> 
> > This diff adds sizes to free(), which completes ufs/ffs.
> 
> It's broken at least for softdep+UFS2.  This chunk blows up:
> 
> > --- ufs/ffs/ffs_softdep.c   10 Feb 2018 05:24:23 -  1.138
> > +++ ufs/ffs/ffs_softdep.c   29 Mar 2018 02:55:37 -
> > @@ -4034,7 +4036,8 @@ handle_written_inodeblock(struct inodede
> > *dp1 = *inodedep->id_savedino1;
> > else
> > *dp2 = *inodedep->id_savedino2;
> > -   free(inodedep->id_savedino1, M_INODEDEP, 0);
> > +   free(inodedep->id_savedino1, M_INODEDEP,
> > +   sizeof(struct ufs1_dinode));
> > inodedep->id_savedino1 = NULL;
> > if ((bp->b_flags & B_DELWRI) == 0)
> > stat_inode_bitmap++;
> 
> panic: free: size too small 128 <= 256 / 2 (0x8099a700) type inodedep
> Stopped at  db_enter+0x5:   popq%rbp
> TIDPIDUID PRFLAGS PFLAGS  CPU  COMMAND
> db_enter() at db_enter+0x5
> panic() at panic+0x129
> free(ff07259d97c8,ff07259d97c8,ff071c6d4ef8) at free+0x35b
> handle_written_inodeblock(ff07259d97c8,ff071c6d4e00) at 
> handle_written_
> inodeblock+0x1b5
> softdep_disk_write_complete(ff071c6d4e00) at 
> softdep_disk_write_complete+0x
> 255
> biodone(ff071bd999e8) at biodone+0x73
> 
> id_savedino1 and id_savedino2 are a union, so it's pretty clear
> that the free() needs to distinguish between the UFS1 and UFS2 cases
> like the preceding lines do.
> 
> I don't know if check_inode_unwritten() needs to make the same
> distinction:
> 
> > @@ -2307,7 +2307,8 @@ check_inode_unwritten(struct inodedep *i
> > if (inodedep->id_state & ONWORKLIST)
> > WORKLIST_REMOVE(>id_list);
> > if (inodedep->id_savedino1 != NULL) {
> > -   free(inodedep->id_savedino1, M_INODEDEP, 0);
> > +   free(inodedep->id_savedino1, M_INODEDEP,
> > +   sizeof(struct ufs1_dinode));
> > inodedep->id_savedino1 = NULL;
> > }
> > if (free_inodedep(inodedep) == 0) {
> 
> -- 
> Christian "naddy" Weisgerber  na...@mips.inka.de
>

Can you try this diff?

Index: ufs/ffs/ffs_softdep.c
===
RCS file: /cvs/src/sys/ufs/ffs/ffs_softdep.c,v
retrieving revision 1.139
diff -u -p -r1.139 ffs_softdep.c
--- ufs/ffs/ffs_softdep.c   30 Mar 2018 17:35:20 -  1.139
+++ ufs/ffs/ffs_softdep.c   1 Apr 2018 00:55:10 -
@@ -2291,6 +2291,8 @@ softdep_freefile(struct vnode *pvp, ufsi
 STATIC int
 check_inode_unwritten(struct inodedep *inodedep)
 {
+   size_t freesize;
+
splassert(IPL_BIO);
 
if ((inodedep->id_state & DEPCOMPLETE) != 0 ||
@@ -2307,8 +2309,11 @@ check_inode_unwritten(struct inodedep *i
if (inodedep->id_state & ONWORKLIST)
WORKLIST_REMOVE(>id_list);
if (inodedep->id_savedino1 != NULL) {
-   free(inodedep->id_savedino1, M_INODEDEP,
-   sizeof(struct ufs1_dinode));
+   if (inodedep->id_fs->fs_magic == FS_UFS1_MAGIC)
+   freesize = sizeof(struct ufs1_dinode);
+   else
+   freesize = sizeof(struct ufs2_dinode);
+   free(inodedep->id_savedino1, M_INODEDEP, freesize);
inodedep->id_savedino1 = NULL;
}
if (free_inodedep(inodedep) == 0) {
@@ -4006,6 +4011,7 @@ handle_written_inodeblock(struct inodede
struct ufs1_dinode *dp1 = NULL;
struct ufs2_dinode *dp2 = NULL;
int hadchanges, fstype;
+   size_t freesize;
 
splassert(IPL_BIO);
 
@@ -4015,10 +4021,12 @@ handle_written_inodeblock(struct inodede
 
if (inodedep->id_fs->fs_magic == FS_UFS1_MAGIC) {
fstype = UM_UFS1;
+   freesize = sizeof(struct ufs1_dinode);
dp1 = (struct ufs1_dinode *) bp->b_data +
ino_to_fsbo(inodedep->id_fs, inodedep->id_ino);
} else {
fstype = UM_UFS2;
+   freesize = sizeof(struct ufs2_dinode);
dp2 = (struct ufs2_dinode *) bp->b_data +
ino_to_fsbo(inodedep->id_fs, inodedep->id_ino);
}
@@ -4035,8 +4043,7 @@ handle_written_inodeblock(struct inodede
*dp1 = *inodedep->id_savedino1;
else
*dp2 = *inodedep->id_savedino2;
-   free(inodedep->id_savedino1, M_INODEDEP,
-   sizeof(struct ufs1_dinode));
+   free(inodedep->id_savedino1, M_INODEDEP, freesize);
inodedep->id_savedino1 = NULL;
if ((bp->b_flags & B_DELWRI) == 0)
stat_inode_bitmap++;

arch/amd64 free

[David Hill]

Hello -

This diff is more involved.

In est_acpi_pss_changed, a new table is allocated but n isn't updated,
which keep tracks of the number allocated.  Set it.

In est_init, fake_table was being allocated with 3.  Change that to
allocate exactly what is need by setting n earlier.


Regarding the isa/ section, in _isa_bus_dmamap_create, cookiesize is:
  cookiesize = sizeof(struct isa_dma_cookie);
  ...
  if ((avail_end > ISA_DMA_BOUNCE_THRESHOLD &&
  (flags & ISABUS_DMA_32BIT) == 0) ||
  ((map->_dm_size / NBPG) + 1) > map->_dm_segcnt) {
cookieflags |= ID_MIGHT_NEED_BOUNCE;
cookiesize += (sizeof(bus_dma_segment_t) * map->_dm_segcnt);
  }

so I am assuming if ID_MIGHT_NEED_BOUNCE is set, that cookiesize gets
they larger size.

OK?

Index: arch/amd64/amd64/est.c
===
RCS file: /cvs/src/sys/arch/amd64/amd64/est.c,v
retrieving revision 1.39
diff -u -p -r1.39 est.c
--- arch/amd64/amd64/est.c  6 Mar 2016 22:41:24 -   1.39
+++ arch/amd64/amd64/est.c  31 Mar 2018 19:11:48 -
@@ -284,6 +284,7 @@ est_acpi_pss_changed(struct acpicpu_pss 
free(acpilist, M_DEVBUF, sizeof(struct fqlist));
return;
}
+   acpilist->n = npss;
 
for (i = 0; i < npss; i++) {
acpilist->table[i].mhz = pss[i].pss_core_freq;
@@ -377,10 +378,13 @@ est_init(struct cpu_info *ci)
"list\n", cpu_device);
return;
}
+   if (cur == idhi || cur == idlo)
+   fake_fqlist->n = 2; 
+   else
+   fake_fqlist->n = 3;
 
-
-   if ((fake_table = mallocarray(3, sizeof(struct est_op),
-   M_DEVBUF, M_NOWAIT)) == NULL) {
+   if ((fake_table = mallocarray(fake_fqlist->n,
+   sizeof(struct est_op), M_DEVBUF, M_NOWAIT)) == NULL) {
free(fake_fqlist, M_DEVBUF, sizeof(struct fqlist));
printf("%s: EST: cannot allocate memory for fake "
"table\n", cpu_device);
@@ -388,13 +392,12 @@ est_init(struct cpu_info *ci)
}
fake_table[0].ctrl = idhi;
fake_table[0].mhz = MSR2MHZ(idhi, bus_clock);
-   if (cur == idhi || cur == idlo) {
+   if (fake_fqlist->n == 2) {
printf("%s: using only highest and lowest power "
   "states\n", cpu_device);
 
fake_table[1].ctrl = idlo;
fake_table[1].mhz = MSR2MHZ(idlo, bus_clock);
-   fake_fqlist->n = 2;
} else {
printf("%s: using only highest, current and lowest "
"power states\n", cpu_device);
@@ -404,7 +407,6 @@ est_init(struct cpu_info *ci)
 
fake_table[2].ctrl = idlo;
fake_table[2].mhz = MSR2MHZ(idlo, bus_clock);
-   fake_fqlist->n = 3;
}
 
fake_fqlist->vendor = vendor;
@@ -441,7 +443,7 @@ est_init(struct cpu_info *ci)
return;
 
 nospeedstep:
-   free(est_fqlist->table, M_DEVBUF, 0);
+   free(est_fqlist->table, M_DEVBUF, est_fqlist->n * sizeof(struct 
est_op));
free(est_fqlist, M_DEVBUF, sizeof(*est_fqlist));
 }
 
Index: arch/amd64/isa/isa_machdep.c
===
RCS file: /cvs/src/sys/arch/amd64/isa/isa_machdep.c,v
retrieving revision 1.29
diff -u -p -r1.29 isa_machdep.c
--- arch/amd64/isa/isa_machdep.c14 Oct 2017 04:44:43 -  1.29
+++ arch/amd64/isa/isa_machdep.c31 Mar 2018 19:11:48 -
@@ -455,6 +455,7 @@ void
 _isa_bus_dmamap_destroy(bus_dma_tag_t t, bus_dmamap_t map)
 {
struct isa_dma_cookie *cookie = map->_dm_cookie;
+   size_t cookiesize = sizeof(struct isa_dma_cookie);
 
/*
 * Free any bounce pages this map might hold.
@@ -462,7 +463,10 @@ _isa_bus_dmamap_destroy(bus_dma_tag_t t,
if (cookie->id_flags & ID_HAS_BOUNCE)
_isa_dma_free_bouncebuf(t, map);
 
-   free(cookie, M_DEVBUF, 0);
+   if (cookie->id_flags & ID_MIGHT_NEED_BOUNCE)
+   cookiesize += (sizeof(bus_dma_segment_t) * map->_dm_segcnt);
+
+   free(cookie, M_DEVBUF, cookiesize);
_bus_dmamap_destroy(t, map);
 }
 

mfs free

[David Hill]

Add the free size. (allocated in mfs_vfsops.c)

mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK | M_ZERO);
devvp->v_data = mfsp;

OK?

Index: ufs/mfs/mfs_vnops.c
===
RCS file: /cvs/src/sys/ufs/mfs/mfs_vnops.c,v
retrieving revision 1.49
diff -u -p -r1.49 mfs_vnops.c
--- ufs/mfs/mfs_vnops.c 7 Nov 2016 00:26:33 -   1.49
+++ ufs/mfs/mfs_vnops.c 31 Mar 2018 19:11:58 -
@@ -238,7 +238,7 @@ mfs_reclaim(void *v)
struct vop_reclaim_args *ap = v;
struct vnode *vp = ap->a_vp;
 
-   free(vp->v_data, M_MFSNODE, 0);
+   free(vp->v_data, M_MFSNODE, sizeof(struct mfsnode));
vp->v_data = NULL;
return (0);
 }

netinet memcpy/free

[David Hill]

Hello -

memcpy can be used on freshly allocated memory.  Fill in the free size
for it.

OK?

Index: netinet/tcp_subr.c
===
RCS file: /cvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.169
diff -u -p -r1.169 tcp_subr.c
--- netinet/tcp_subr.c  18 Mar 2018 21:05:21 -  1.169
+++ netinet/tcp_subr.c  31 Mar 2018 19:12:04 -
@@ -952,7 +952,7 @@ tcp_signature_tdb_init(struct tdb *tdbp,
tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_NOWAIT);
if (tdbp->tdb_amxkey == NULL)
return (ENOMEM);
-   bcopy(ii->ii_authkey, tdbp->tdb_amxkey, ii->ii_authkeylen);
+   memcpy(tdbp->tdb_amxkey, ii->ii_authkey, ii->ii_authkeylen);
tdbp->tdb_amxkeylen = ii->ii_authkeylen;
 
return (0);
@@ -963,7 +963,7 @@ tcp_signature_tdb_zeroize(struct tdb *td
 {
if (tdbp->tdb_amxkey) {
explicit_bzero(tdbp->tdb_amxkey, tdbp->tdb_amxkeylen);
-   free(tdbp->tdb_amxkey, M_XDATA, 0);
+   free(tdbp->tdb_amxkey, M_XDATA, tdbp->tdb_amxkeylen);
tdbp->tdb_amxkey = NULL;
}
 

ufs free()

[David Hill]

Hello -

This diff adds sizes to free(), which completes ufs/ffs.

OK?

Index: ufs/ffs/ffs_inode.c
===
RCS file: /cvs/src/sys/ufs/ffs/ffs_inode.c,v
retrieving revision 1.76
diff -u -p -r1.76 ffs_inode.c
--- ufs/ffs/ffs_inode.c 27 Feb 2016 18:50:38 -  1.76
+++ ufs/ffs/ffs_inode.c 29 Mar 2018 02:55:36 -
@@ -560,7 +560,7 @@ ffs_indirtrunc(struct inode *ip, daddr_t
}
}
if (copy != NULL) {
-   free(copy, M_TEMP, 0);
+   free(copy, M_TEMP, fs->fs_bsize);
} else {
bp->b_flags |= B_INVAL;
brelse(bp);
Index: ufs/ffs/ffs_softdep.c
===
RCS file: /cvs/src/sys/ufs/ffs/ffs_softdep.c,v
retrieving revision 1.138
diff -u -p -r1.138 ffs_softdep.c
--- ufs/ffs/ffs_softdep.c   10 Feb 2018 05:24:23 -  1.138
+++ ufs/ffs/ffs_softdep.c   29 Mar 2018 02:55:37 -
@@ -2307,7 +2307,8 @@ check_inode_unwritten(struct inodedep *i
if (inodedep->id_state & ONWORKLIST)
WORKLIST_REMOVE(>id_list);
if (inodedep->id_savedino1 != NULL) {
-   free(inodedep->id_savedino1, M_INODEDEP, 0);
+   free(inodedep->id_savedino1, M_INODEDEP,
+   sizeof(struct ufs1_dinode));
inodedep->id_savedino1 = NULL;
}
if (free_inodedep(inodedep) == 0) {
@@ -3845,7 +3846,8 @@ softdep_disk_write_complete(struct buf *
if (indirdep->ir_state & GOINGAWAY)
panic("disk_write_complete: indirdep gone");
memcpy(bp->b_data, indirdep->ir_saveddata, 
bp->b_bcount);
-   free(indirdep->ir_saveddata, M_INDIRDEP, 0);
+   free(indirdep->ir_saveddata, M_INDIRDEP,
+   sizeof(struct ufs1_dinode));
indirdep->ir_saveddata = NULL;
indirdep->ir_state &= ~UNDONE;
indirdep->ir_state |= ATTACHED;
@@ -4034,7 +4036,8 @@ handle_written_inodeblock(struct inodede
*dp1 = *inodedep->id_savedino1;
else
*dp2 = *inodedep->id_savedino2;
-   free(inodedep->id_savedino1, M_INODEDEP, 0);
+   free(inodedep->id_savedino1, M_INODEDEP,
+   sizeof(struct ufs1_dinode));
inodedep->id_savedino1 = NULL;
if ((bp->b_flags & B_DELWRI) == 0)
stat_inode_bitmap++;
Index: ufs/ffs/ffs_vfsops.c
===
RCS file: /cvs/src/sys/ufs/ffs/ffs_vfsops.c,v
retrieving revision 1.173
diff -u -p -r1.173 ffs_vfsops.c
--- ufs/ffs/ffs_vfsops.c15 Mar 2018 04:22:16 -  1.173
+++ ufs/ffs/ffs_vfsops.c29 Mar 2018 02:55:37 -
@@ -447,7 +447,7 @@ success:
fs->fs_clean = ronly &&
(fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
if (ronly)
-   free(fs->fs_contigdirs, M_UFSMNT, 0);
+   free(fs->fs_contigdirs, M_UFSMNT, fs->fs_ncg);
}
if (!ronly) {
if (mp->mnt_flag & MNT_SOFTDEP)
@@ -837,7 +837,7 @@ ffs_mountfs(struct vnode *devvp, struct 
size = (blks - i) * fs->fs_fsize;
error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, );
if (error) {
-   free(fs->fs_csp, M_UFSMNT, 0);
+   free(fs->fs_csp, M_UFSMNT, fs->fs_cssize);
goto out;
}
memcpy(space, bp->b_data, size);
@@ -910,8 +910,8 @@ ffs_mountfs(struct vnode *devvp, struct 
if (ronly == 0) {
if ((fs->fs_flags & FS_DOSOFTDEP) &&
(error = softdep_mount(devvp, mp, fs, cred)) != 0) {
-   free(fs->fs_csp, M_UFSMNT, 0);
-   free(fs->fs_contigdirs, M_UFSMNT, 0);
+   free(fs->fs_csp, M_UFSMNT, fs->fs_cssize);
+   free(fs->fs_contigdirs, M_UFSMNT, fs->fs_ncg);
goto out;
}
fs->fs_fmod = 1;
@@ -1046,7 +1046,7 @@ ffs_unmount(struct mount *mp, int mntfla
fs->fs_clean = 0;
return (error);
}
-   free(fs->fs_contigdirs, M_UFSMNT, 0);
+   free(fs->fs_contigdirs, M_UFSMNT, fs->fs_ncg);
}
ump->um_devvp->v_specmountpoint = NULL;
 
@@ -1055,7 +1055,7 @@ ffs_unmount(struct mount *mp, int mntfla
(void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
NOCRED, p);
vput(ump->um_devvp);
-   free(fs->fs_csp, M_UFSMNT, 0);
+   free(fs->fs_csp, M_UFSMNT, 

Re: use hashfree on pcb hash tables

[David Hill]

round #2.

keep track of size (num of elements) in the inpcbtable struct.
passes regress tests.

OK?

Index: netinet/in_pcb.c
===
RCS file: /cvs/src/sys/netinet/in_pcb.c,v
retrieving revision 1.228
diff -u -p -r1.228 in_pcb.c
--- netinet/in_pcb.c19 Feb 2018 08:59:53 -  1.228
+++ netinet/in_pcb.c29 Mar 2018 02:21:34 -
@@ -206,6 +206,7 @@ in_pcbinit(struct inpcbtable *table, int
if (table->inpt_lhashtbl == NULL)
panic("in_pcbinit: hashinit failed for lport");
table->inpt_count = 0;
+   table->inpt_size = hashsize;
arc4random_buf(>inpt_key, sizeof(table->inpt_key));
 }
 
@@ -998,32 +999,34 @@ int
 in_pcbresize(struct inpcbtable *table, int hashsize)
 {
u_long nhash, nlhash;
+   int osize;
void *nhashtbl, *nlhashtbl, *ohashtbl, *olhashtbl;
struct inpcb *inp0, *inp1;
 
ohashtbl = table->inpt_hashtbl;
olhashtbl = table->inpt_lhashtbl;
+   osize = table->inpt_size;
 
nhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT, );
+   if (nhashtbl == NULL)
+   return ENOBUFS;
nlhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT, );
-   if (nhashtbl == NULL || nlhashtbl == NULL) {
-   if (nhashtbl != NULL)
-   free(nhashtbl, M_PCB, 0);
-   if (nlhashtbl != NULL)
-   free(nlhashtbl, M_PCB, 0);
-   return (ENOBUFS);
+   if (nlhashtbl == NULL) {
+   hashfree(nhashtbl, hashsize, M_PCB);
+   return ENOBUFS;
}
table->inpt_hashtbl = nhashtbl;
table->inpt_lhashtbl = nlhashtbl;
table->inpt_hash = nhash;
table->inpt_lhash = nlhash;
+   table->inpt_size = hashsize;
arc4random_buf(>inpt_key, sizeof(table->inpt_key));
 
TAILQ_FOREACH_SAFE(inp0, >inpt_queue, inp_queue, inp1) {
in_pcbrehash(inp0);
}
-   free(ohashtbl, M_PCB, 0);
-   free(olhashtbl, M_PCB, 0);
+   hashfree(ohashtbl, osize, M_PCB);
+   hashfree(olhashtbl, osize, M_PCB);
 
return (0);
 }
Index: netinet/in_pcb.h
===
RCS file: /cvs/src/sys/netinet/in_pcb.h,v
retrieving revision 1.106
diff -u -p -r1.106 in_pcb.h
--- netinet/in_pcb.h1 Dec 2017 10:33:33 -   1.106
+++ netinet/in_pcb.h29 Mar 2018 02:21:34 -
@@ -152,7 +152,7 @@ struct inpcbtable {
struct inpcbhead *inpt_hashtbl, *inpt_lhashtbl;
SIPHASH_KEY inpt_key;
u_longinpt_hash, inpt_lhash;
-   int   inpt_count;
+   int   inpt_count, inpt_size;
 };
 
 /* flags in inp_flags: */

Re: use hashfree on pcb hash tables

[David Hill]

Ignore for now.

A lovely regress/ test exposes a problem in my diff. :)

On Tue, Mar 27, 2018 at 01:11:19PM -0400, David Hill wrote:
> Hello -
> 
> The hash tables are allocated with hashinit, so free them with hashfree.
> This gives the bonus of calling free() with a size as well.
> 

use hashfree on pcb hash tables

[David Hill]

Hello -

The hash tables are allocated with hashinit, so free them with hashfree.
This gives the bonus of calling free() with a size as well.

OK?

Index: in_pcb.c
===
RCS file: /cvs/src/sys/netinet/in_pcb.c,v
retrieving revision 1.228
diff -u -p -r1.228 in_pcb.c
--- in_pcb.c19 Feb 2018 08:59:53 -  1.228
+++ in_pcb.c27 Mar 2018 17:06:51 -
@@ -997,20 +997,21 @@ in_pcbrehash(struct inpcb *inp)
 int
 in_pcbresize(struct inpcbtable *table, int hashsize)
 {
-   u_long nhash, nlhash;
+   u_long nhash, nlhash, ohash, olhash;
void *nhashtbl, *nlhashtbl, *ohashtbl, *olhashtbl;
struct inpcb *inp0, *inp1;
 
ohashtbl = table->inpt_hashtbl;
+   ohash = table->inpt_hash;
olhashtbl = table->inpt_lhashtbl;
+   olhash = table->inpt_lhash;
 
nhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT, );
+   if (nhashtbl == NULL)
+   return (ENOBUFS);
nlhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT, );
-   if (nhashtbl == NULL || nlhashtbl == NULL) {
-   if (nhashtbl != NULL)
-   free(nhashtbl, M_PCB, 0);
-   if (nlhashtbl != NULL)
-   free(nlhashtbl, M_PCB, 0);
+   if (nlhashtbl == NULL) {
+   hashfree(nhashtbl, nhash, M_PCB);
return (ENOBUFS);
}
table->inpt_hashtbl = nhashtbl;
@@ -1022,8 +1023,8 @@ in_pcbresize(struct inpcbtable *table, i
TAILQ_FOREACH_SAFE(inp0, >inpt_queue, inp_queue, inp1) {
in_pcbrehash(inp0);
}
-   free(ohashtbl, M_PCB, 0);
-   free(olhashtbl, M_PCB, 0);
+   hashfree(ohashtbl, ohash, M_PCB);
+   hashfree(olhashtbl, olhash, M_PCB);
 
return (0);
 }

netinet bcopy -> memcpy/memmove

[David Hill]

Hello -

A few bcopy conversions to memcpy where the memory does not overlap,
otherwise memmove. 

OK?

Index: netinet/ip_ah.c
===
RCS file: /cvs/src/sys/netinet/ip_ah.c,v
retrieving revision 1.138
diff -u -p -r1.138 ip_ah.c
--- netinet/ip_ah.c 14 Mar 2018 22:38:46 -  1.138
+++ netinet/ip_ah.c 27 Mar 2018 16:10:03 -
@@ -899,8 +899,8 @@ ah_input_cb(struct cryptop *crp)
 * mbuf...do an overlapping copy of the
 * remainder of the mbuf over the ESP header.
 */
-   bcopy(mtod(m1, u_char *) + roff + rplen +
-   ahx->authsize, mtod(m1, u_char *) + roff,
+   memmove(mtod(m1, u_char *) + roff,
+   mtod(m1, u_char *) + roff + rplen + ahx->authsize,
m1->m_len - (roff + rplen + ahx->authsize));
m1->m_len -= rplen + ahx->authsize;
m->m_pkthdr.len -= rplen + ahx->authsize;
Index: netinet/tcp_subr.c
===
RCS file: /cvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.169
diff -u -p -r1.169 tcp_subr.c
--- netinet/tcp_subr.c  18 Mar 2018 21:05:21 -  1.169
+++ netinet/tcp_subr.c  27 Mar 2018 16:10:03 -
@@ -328,11 +328,11 @@ tcp_respond(struct tcpcb *tp, caddr_t te
th = (struct tcphdr *)(ip6 + 1);
tlen = sizeof(*ip6) + sizeof(*th);
if (th0) {
-   bcopy(template, ip6, sizeof(*ip6));
-   bcopy(th0, th, sizeof(*th));
+   memcpy(ip6, template, sizeof(*ip6));
+   memcpy(th, th0, sizeof(*th));
xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
} else {
-   bcopy(template, ip6, tlen);
+   memcpy(ip6, template, tlen);
}
break;
 #endif /* INET6 */
@@ -341,11 +341,11 @@ tcp_respond(struct tcpcb *tp, caddr_t te
th = (struct tcphdr *)(ip + 1);
tlen = sizeof(*ip) + sizeof(*th);
if (th0) {
-   bcopy(template, ip, sizeof(*ip));
-   bcopy(th0, th, sizeof(*th));
+   memcpy(ip, template, sizeof(*ip));
+   memcpy(th, th0, sizeof(*th));
xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, u_int32_t);
} else {
-   bcopy(template, ip, tlen);
+   memcpy(ip, template, tlen);
}
break;
}
@@ -952,7 +952,7 @@ tcp_signature_tdb_init(struct tdb *tdbp,
tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_NOWAIT);
if (tdbp->tdb_amxkey == NULL)
return (ENOMEM);
-   bcopy(ii->ii_authkey, tdbp->tdb_amxkey, ii->ii_authkeylen);
+   memcpy(tdbp->tdb_amxkey, ii->ii_authkey, ii->ii_authkeylen);
tdbp->tdb_amxkeylen = ii->ii_authkeylen;
 
return (0);
Index: netinet/udp_usrreq.c
===
RCS file: /cvs/src/sys/netinet/udp_usrreq.c,v
retrieving revision 1.245
diff -u -p -r1.245 udp_usrreq.c
--- netinet/udp_usrreq.c1 Dec 2017 10:33:33 -   1.245
+++ netinet/udp_usrreq.c27 Mar 2018 16:10:03 -
@@ -295,8 +295,8 @@ udp_input(struct mbuf **mp, int *offp, i
}
 
/* remove the UDP header */
-   bcopy(mtod(m, u_char *),
-   mtod(m, u_char *) + sizeof(struct udphdr), iphlen);
+   memmove(mtod(m, u_char *) + sizeof(struct udphdr),
+   mtod(m, u_char *), iphlen);
m_adj(m, sizeof(struct udphdr));
skip -= sizeof(struct udphdr);
 

netinet6 - bcopy -> memcpy

[David Hill]

Hello -

A few bcopy to memcpy conversions where the memory does not overlap.

OK?

Index: netinet6/icmp6.c
===
RCS file: /cvs/src/sys/netinet6/icmp6.c,v
retrieving revision 1.221
diff -u -p -r1.221 icmp6.c
--- netinet6/icmp6.c14 Dec 2017 14:26:50 -  1.221
+++ netinet6/icmp6.c13 Mar 2018 21:32:32 -
@@ -1075,7 +1075,7 @@ icmp6_reflect(struct mbuf *m, size_t off
if ((m = m_pullup(m, l)) == NULL)
return;
}
-   bcopy((caddr_t), mtod(m, caddr_t), sizeof(nip6));
+   memcpy(mtod(m, caddr_t), (caddr_t), sizeof(nip6));
} else /* off == sizeof(struct ip6_hdr) */ {
size_t l;
l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
@@ -1268,7 +1268,7 @@ icmp6_redirect_input(struct mbuf *m, int
bzero(, sizeof(sin6));
sin6.sin6_family = AF_INET6;
sin6.sin6_len = sizeof(struct sockaddr_in6);
-   bcopy(, _addr, sizeof(reddst6));
+   memcpy(_addr, , sizeof(reddst6));
rt = rtalloc(sin6tosa(), 0, m->m_pkthdr.ph_rtableid);
if (rt) {
if (rt->rt_gateway == NULL ||
@@ -1376,9 +1376,9 @@ icmp6_redirect_input(struct mbuf *m, int
sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = 
AF_INET6;
sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
sizeof(struct sockaddr_in6);
-   bcopy(, _addr, sizeof(struct in6_addr));
-   bcopy(, _addr, sizeof(struct in6_addr));
-   bcopy(, _addr, sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
rtredirect(sin6tosa(), sin6tosa(), sin6tosa(),
, m->m_pkthdr.ph_rtableid);
 
@@ -1395,7 +1395,7 @@ icmp6_redirect_input(struct mbuf *m, int
bzero(, sizeof(sdst));
sdst.sin6_family = AF_INET6;
sdst.sin6_len = sizeof(struct sockaddr_in6);
-   bcopy(, _addr, sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
pfctlinput(PRC_REDIRECT_HOST, sin6tosa());
}
 
Index: netinet6/in6_ifattach.c
===
RCS file: /cvs/src/sys/netinet6/in6_ifattach.c,v
retrieving revision 1.106
diff -u -p -r1.106 in6_ifattach.c
--- netinet6/in6_ifattach.c 13 Mar 2018 13:58:03 -  1.106
+++ netinet6/in6_ifattach.c 13 Mar 2018 21:32:33 -
@@ -165,7 +165,7 @@ in6_get_hw_ifid(struct ifnet *ifp, struc
 
/* make EUI64 address */
if (addrlen == 8)
-   bcopy(addr, >s6_addr[8], 8);
+   memcpy(>s6_addr[8], addr, 8);
else if (addrlen == 6) {
in6->s6_addr[8] = addr[0];
in6->s6_addr[9] = addr[1];
@@ -244,7 +244,7 @@ in6_get_soii_ifid(struct ifnet *ifp, str
SHA512Update(, ip6_soiikey, sizeof(ip6_soiikey));
SHA512Final(digest, );
 
-   bcopy(digest + (sizeof(digest) - 8), >s6_addr[8], 8);
+   memcpy(>s6_addr[8], digest + (sizeof(digest) - 8), 8);
 
return 0;
 }
@@ -464,7 +464,7 @@ in6_nigroup(struct ifnet *ifp, const cha
sa6->sin6_addr.s6_addr16[0] = htons(0xff02);
sa6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
sa6->sin6_addr.s6_addr8[11] = 2;
-   bcopy(digest, >sin6_addr.s6_addr32[3],
+   memcpy(>sin6_addr.s6_addr32[3], digest,
sizeof(sa6->sin6_addr.s6_addr32[3]));
 
return 0;
Index: netinet6/ip6_output.c
===
RCS file: /cvs/src/sys/netinet6/ip6_output.c,v
retrieving revision 1.234
diff -u -p -r1.234 ip6_output.c
--- netinet6/ip6_output.c   19 Feb 2018 08:59:53 -  1.234
+++ netinet6/ip6_output.c   13 Mar 2018 21:32:33 -
@@ -850,7 +850,7 @@ ip6_copyexthdr(struct mbuf **mp, caddr_t
}
m->m_len = hlen;
if (hdr)
-   bcopy(hdr, mtod(m, caddr_t), hlen);
+   memcpy(mtod(m, caddr_t), hdr, hlen);
 
*mp = m;
return (0);
@@ -918,7 +918,7 @@ ip6_insert_jumboopt(struct ip6_exthdrs *
if (!n)
return (ENOBUFS);
n->m_len = oldoptlen + JUMBOOPTLEN;
-   bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
+   memcpy(mtod(n, caddr_t), mtod(mopt, caddr_t),
  oldoptlen);
optbuf = mtod(n, u_int8_t *) + oldoptlen;
m_freem(mopt);

Re: tcp timeout milliseconds

[David Hill]

OK dhill@

On Tue, Feb 06, 2018 at 05:10:17PM +0100, Alexander Bluhm wrote:
> Hi,
> 
> Historically TCP timeouts were implemented with pr_slowtimo and
> pr_fasttimo.  That is the reason why we have two timeout mechanisms
> with complicated ticks calculation.
> 
> I would like to move to milliseconds and merge them eventually.
> This makes it easier to see the actual values.
> 
> Let's get rid of some easy ticks and hz.
> 
> ok?
> 
> bluhm
> 

Re: pf statekey inp assertion failed

[David Hill]

On Wed, Jun 21, 2017 at 01:21:50AM +0200, Alexander Bluhm wrote:
> Hi,
> 
> I saw a crash on an OpenBSD 6.1 based system when a kassert in
> pf_state_key_unref() was triggert.
> 
> kernel diagnostic assertion "(sk->inp == NULL) || (sk->inp->inp_pf_sk == 
> NULL)" 
>  failed: file "../../../../../net/pf.c", line 7155
>   
> 
> panic() at panic+0xfe 
>   
> __assert() at __assert+0x25   
>   
> pf_state_key_unref() at pf_state_key_unref+0xc6   
>   
> pf_pkt_unlink_state_key() at pf_pkt_unlink_state_key+0x15 
>   
> m_free() at m_free+0xc0   
>   
> soreceive() at soreceive+0xb5d
>   
> recvit() at recvit+0x13a  
>   
> sys_recvmsg() at sys_recvmsg+0x107
>   
> syscall() at syscall+0x2df
>   
> 
> The problem is that setting the inp pointer in the statekey to NULL
> is delayed until the statekey refcounter reaches 0.  So the inp
> could get linked to another statekey while the mbuf in the socket
> buffer was keeping the refcounter at 1.
> 
> The sk->inp should be set to NULL immediately, then the kassert can
> get even stricter.
> 
> ok?

ok!! dhill@

I hit this too.  See bugs@ archives :)

> 
> bluhm
> 
> Index: net/pf.c
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/net/pf.c,v
> retrieving revision 1.1034
> diff -u -p -r1.1034 pf.c
> --- net/pf.c  5 Jun 2017 22:18:28 -   1.1034
> +++ net/pf.c  20 Jun 2017 22:37:43 -
> @@ -779,6 +779,7 @@ pf_state_key_detach(struct pf_state *s, 
>   sk->removed = 1;
>   pf_state_key_unlink_reverse(sk);
>   pf_inpcb_unlink_state_key(sk->inp);
> + sk->inp = NULL;
>   pf_state_key_unref(sk);
>   }
>  }
> @@ -7147,8 +7148,7 @@ pf_state_key_unref(struct pf_state_key *
>   /* state key must be unlinked from reverse key */
>   KASSERT(sk->reverse == NULL);
>   /* state key must be unlinked from socket */
> - KASSERT((sk->inp == NULL) || (sk->inp->inp_pf_sk == NULL));
> - sk->inp = NULL;
> + KASSERT(sk->inp == NULL);
>   pool_put(_state_key_pl, sk);
>   }
>  }
> 

ip_esp/ip_mroute bcopy->mem*

[David Hill]

Hello -

Convert bcopy to memcpy where the memory does not overlap, otherwise,
use memmove.  While here, change some previous conversions to a simple
assignment.

OK?

Index: ip_esp.c
===
RCS file: /cvs/src/sys/netinet/ip_esp.c,v
retrieving revision 1.146
diff -u -p -r1.146 ip_esp.c
--- ip_esp.c7 Feb 2017 18:18:16 -   1.146
+++ ip_esp.c6 Apr 2017 15:21:27 -
@@ -504,7 +504,7 @@ esp_input(struct mbuf *m, struct tdb *td
tc->tc_spi = tdb->tdb_spi;
tc->tc_proto = tdb->tdb_sproto;
tc->tc_rdomain = tdb->tdb_rdomain;
-   memcpy(>tc_dst, >tdb_dst, sizeof(union sockaddr_union));
+   tc->tc_dst = tdb->tdb_dst;
 
/* Decryption descriptor */
if (espx) {
@@ -707,8 +707,9 @@ esp_input_cb(struct cryptop *crp)
 * overlapping copy of the remainder of the mbuf over the ESP
 * header.
 */
-   bcopy(mtod(m1, u_char *) + roff + hlen,
-   mtod(m1, u_char *) + roff, m1->m_len - (roff + hlen));
+   memmove(mtod(m1, u_char *) + roff, 
+   mtod(m1, u_char *) + roff + hlen,
+   m1->m_len - (roff + hlen));
m1->m_len -= hlen;
m->m_pkthdr.len -= hlen;
}
@@ -910,7 +911,7 @@ esp_output(struct mbuf *m, struct tdb *t
}
 
/* Initialize ESP header. */
-   bcopy((caddr_t) >tdb_spi, mtod(mo, caddr_t) + roff,
+   memcpy(mtod(mo, caddr_t) + roff, (caddr_t) >tdb_spi,
sizeof(u_int32_t));
tdb->tdb_rpl++;
replay = htonl((u_int32_t)tdb->tdb_rpl);
@@ -992,7 +993,7 @@ esp_output(struct mbuf *m, struct tdb *t
tc->tc_spi = tdb->tdb_spi;
tc->tc_proto = tdb->tdb_sproto;
tc->tc_rdomain = tdb->tdb_rdomain;
-   memcpy(>tc_dst, >tdb_dst, sizeof(union sockaddr_union));
+   tc->tc_dst = tdb->tdb_dst;
 
/* Crypto operation descriptor. */
crp->crp_ilen = m->m_pkthdr.len; /* Total input length. */
Index: ip_mroute.c
===
RCS file: /cvs/src/sys/netinet/ip_mroute.c,v
retrieving revision 1.112
diff -u -p -r1.112 ip_mroute.c
--- ip_mroute.c 17 Mar 2017 14:59:29 -  1.112
+++ ip_mroute.c 6 Apr 2017 15:21:28 -
@@ -951,10 +951,10 @@ add_mfc(struct socket *so, struct mbuf *
 */
if (mrt_api_config & MRT_API_FLAGS_ALL) {
struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *);
-   bcopy(mp2, (caddr_t), sizeof(*mp2));
+   memcpy((caddr_t), mp2, sizeof(*mp2));
} else {
struct mfcctl *mp = mtod(m, struct mfcctl *);
-   bcopy(mp, (caddr_t), sizeof(*mp));
+   memcpy((caddr_t), mp, sizeof(*mp));
memset((caddr_t) + sizeof(struct mfcctl), 0,
sizeof(mfcctl2) - sizeof(struct mfcctl));
}
@@ -988,7 +988,7 @@ del_mfc(struct socket *so, struct mbuf *
 
mp = mtod(m, struct mfcctl *);
 
-   bcopy(mp, (caddr_t), sizeof(*mp));
+   memcpy((caddr_t), mp, sizeof(*mp));
memset((caddr_t) + sizeof(struct mfcctl), 0,
sizeof(mfcctl2) - sizeof(struct mfcctl));
 

ip6_input.c and mld6.c bcopy -> memcpy

[David Hill]

Hello -

The following diff removes bcopy from ip6_input.c and mld6.c by using
memcpy as the memory does not overlap.

OK?

Index: ip6_input.c
===
RCS file: /cvs/src/sys/netinet6/ip6_input.c,v
retrieving revision 1.181
diff -u -p -r1.181 ip6_input.c
--- ip6_input.c 6 Mar 2017 08:59:07 -   1.181
+++ ip6_input.c 6 Apr 2017 00:50:46 -
@@ -777,7 +777,7 @@ ip6_process_hopopts(struct mbuf *m, u_in
return (-1);
}
optlen = IP6OPT_RTALERT_LEN;
-   bcopy((caddr_t)(opt + 2), (caddr_t)_val, 2);
+   memcpy((caddr_t)_val, (caddr_t)(opt + 2), 2);
*rtalertp = ntohs(rtalert_val);
break;
case IP6OPT_JUMBO:
@@ -810,9 +810,9 @@ ip6_process_hopopts(struct mbuf *m, u_in
 
/*
 * We may see jumbolen in unaligned location, so
-* we'd need to perform bcopy().
+* we'd need to perform memcpy().
 */
-   bcopy(opt + 2, , sizeof(jumboplen));
+   memcpy(, opt + 2, sizeof(jumboplen));
jumboplen = (u_int32_t)htonl(jumboplen);
 
 #if 1
@@ -935,7 +935,7 @@ ip6_savecontrol(struct inpcb *in6p, stru
/* RFC 2292 sec. 5 */
if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
struct in6_pktinfo pi6;
-   bcopy(>ip6_dst, _addr, sizeof(struct in6_addr));
+   memcpy(_addr, >ip6_dst, sizeof(struct in6_addr));
if (IN6_IS_SCOPE_EMBED(_addr))
pi6.ipi6_addr.s6_addr16[1] = 0;
pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
Index: mld6.c
===
RCS file: /cvs/src/sys/netinet6/mld6.c,v
retrieving revision 1.50
diff -u -p -r1.50 mld6.c
--- mld6.c  9 Feb 2017 15:23:35 -   1.50
+++ mld6.c  6 Apr 2017 00:50:46 -
@@ -109,7 +109,7 @@ mld6_init(void)
hbh_buf[3] = 0;
hbh_buf[4] = IP6OPT_ROUTER_ALERT;
hbh_buf[5] = IP6OPT_RTALERT_LEN - 2;
-   bcopy((caddr_t)_code, _buf[6], sizeof(u_int16_t));
+   memcpy(_buf[6], (caddr_t)_code, sizeof(u_int16_t));
 
ip6_opts.ip6po_hbh = hbh;
 }

ipcomp bcopy -> mem*

[David Hill]

Hello -

This diff converts the existing three bcopy's to memcpy or memmove.
The memcpy's are on freshly malloc'd memory so no overlap.

OK?
 
Index: netinet/ip_ipcomp.c
===
RCS file: /cvs/src/sys/netinet/ip_ipcomp.c,v
retrieving revision 1.55
diff -u -p -r1.55 ip_ipcomp.c
--- netinet/ip_ipcomp.c 17 Feb 2017 14:49:03 -  1.55
+++ netinet/ip_ipcomp.c 24 Mar 2017 16:54:37 -
@@ -181,7 +181,7 @@ ipcomp_input(struct mbuf *m, struct tdb 
tc->tc_spi = tdb->tdb_spi;
tc->tc_proto = IPPROTO_IPCOMP;
tc->tc_rdomain = tdb->tdb_rdomain;
-   bcopy(>tdb_dst, >tc_dst, sizeof(union sockaddr_union));
+   memcpy(>tc_dst, >tdb_dst, sizeof(union sockaddr_union));
 
return crypto_dispatch(crp);
 }
@@ -317,8 +317,8 @@ ipcomp_input_cb(struct cryptop *crp)
/* Finally, let's relink */
m1->m_next = mo;
} else {
-   bcopy(mtod(m1, u_char *) + roff + hlen,
-   mtod(m1, u_char *) + roff,
+   memmove(mtod(m1, u_char *) + roff,
+   mtod(m1, u_char *) + roff + hlen,
m1->m_len - (roff + hlen));
m1->m_len -= hlen;
m->m_pkthdr.len -= hlen;
@@ -501,7 +501,7 @@ ipcomp_output(struct mbuf *m, struct tdb
tc->tc_proto = tdb->tdb_sproto;
tc->tc_skip = skip;
tc->tc_rdomain = tdb->tdb_rdomain;
-   bcopy(>tdb_dst, >tc_dst, sizeof(union sockaddr_union));
+   memcpy(>tc_dst, >tdb_dst, sizeof(union sockaddr_union));
 
/* Crypto operation descriptor */
crp->crp_ilen = m->m_pkthdr.len;/* Total input length */

Re: ip_ipip.c / gif(4) percpu counters

[David Hill]

On Thu, Mar 09, 2017 at 03:49:21PM +0100, Jeremie Courreges-Anglas wrote:
> Martin Pieuchot  writes:
> 
> > On 08/03/17(Wed) 12:03, Jeremie Courreges-Anglas wrote:
> >> [...] 
> >> So here's a refreshed diff that initializes the counters directly from
> >> ip_init().  I remove the ipip_init() wrapper to make it clear that
> >> ip_init() is responsible for the job.
> >> 
> >> (Still) ok?
> >
> > I'm against adding more "#if PEUDOIF" in the stack everywhere and it
> > makes no sense to move more globals outside of ip_ipip.c.
> 
> The previous diff was just plain wrong.
> 
> > If you really don't want to use .pr_init, then call ipip_input() from
> > ip_input().
> 
> Actually I'm fine with .pr_init.  There are other protosw entries that
> reach the counters but one shouldn't remove entries carelessly anyway.
> 
> Is this better?
> 
Definitely. OK dhill@
> 
> Index: netinet/in_proto.c
> ===
> RCS file: /d/cvs/src/sys/netinet/in_proto.c,v
> retrieving revision 1.74
> diff -u -p -r1.74 in_proto.c
> --- netinet/in_proto.c2 Mar 2017 08:58:24 -   1.74
> +++ netinet/in_proto.c8 Mar 2017 11:56:45 -
> @@ -239,7 +239,8 @@ struct protosw inetsw[] = {
>.pr_output = rip_output,
>.pr_ctloutput  = rip_ctloutput,
>.pr_usrreq = rip_usrreq,
> -  .pr_sysctl = ipip_sysctl
> +  .pr_sysctl = ipip_sysctl,
> +  .pr_init   = ipip_init
>  },
>  {
>.pr_type   = SOCK_RAW,
> @@ -284,7 +285,8 @@ struct protosw inetsw[] = {
>.pr_output = rip_output,
>.pr_ctloutput  = rip_ctloutput,
>.pr_usrreq = rip_usrreq,
> -  .pr_sysctl = ipip_sysctl
> +  .pr_sysctl = ipip_sysctl,
> +  .pr_init   = ipip_init
>  },
>  #ifdef INET6
>  {
> Index: netinet/ip_ipip.c
> ===
> RCS file: /d/cvs/src/sys/netinet/ip_ipip.c,v
> retrieving revision 1.71
> diff -u -p -r1.71 ip_ipip.c
> --- netinet/ip_ipip.c 29 Jan 2017 19:58:47 -  1.71
> +++ netinet/ip_ipip.c 8 Mar 2017 11:59:04 -
> @@ -84,7 +84,13 @@
>   */
>  int ipip_allow = 0;
>  
> -struct ipipstat ipipstat;
> +struct cpumem *ipipcounters;
> +
> +void
> +ipip_init(void)
> +{
> + ipipcounters = counters_alloc(ipips_ncounters);
> +}
>  
>  /*
>   * Really only a wrapper for ipip_input(), for use with pr_input.
> @@ -95,7 +101,7 @@ ip4_input(struct mbuf **mp, int *offp, i
>   /* If we do not accept IP-in-IP explicitly, drop.  */
>   if (!ipip_allow && ((*mp)->m_flags & (M_AUTH|M_CONF)) == 0) {
>   DPRINTF(("ip4_input(): dropped due to policy\n"));
> - ipipstat.ipips_pdrops++;
> + ipipstat_inc(ipips_pdrops);
>   m_freem(*mp);
>   return IPPROTO_DONE;
>   }
> @@ -129,7 +135,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>   u_int8_t v;
>   sa_family_t af;
>  
> - ipipstat.ipips_ipackets++;
> + ipipstat_inc(ipips_ipackets);
>  
>   m_copydata(m, 0, 1, );
>  
> @@ -143,7 +149,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>   break;
>  #endif
>   default:
> - ipipstat.ipips_family++;
> + ipipstat_inc(ipips_family);
>   m_freem(m);
>   return IPPROTO_DONE;
>   }
> @@ -152,7 +158,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>   if (m->m_len < hlen) {
>   if ((m = m_pullup(m, hlen)) == NULL) {
>   DPRINTF(("ipip_input(): m_pullup() failed\n"));
> - ipipstat.ipips_hdrops++;
> + ipipstat_inc(ipips_hdrops);
>   return IPPROTO_DONE;
>   }
>   }
> @@ -179,7 +185,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>  
>   /* Sanity check */
>   if (m->m_pkthdr.len < sizeof(struct ip)) {
> - ipipstat.ipips_hdrops++;
> + ipipstat_inc(ipips_hdrops);
>   m_freem(m);
>   return IPPROTO_DONE;
>   }
> @@ -195,7 +201,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>   break;
>  #endif
>   default:
> - ipipstat.ipips_family++;
> + ipipstat_inc(ipips_family);
>   m_freem(m);
>   return IPPROTO_DONE;
>   }
> @@ -206,7 +212,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>   if (m->m_len < hlen) {
>   if ((m = m_pullup(m, hlen)) == NULL) {
>   DPRINTF(("ipip_input(): m_pullup() failed\n"));
> - ipipstat.ipips_hdrops++;
> + ipipstat_inc(ipips_hdrops);
>   return IPPROTO_DONE;
>   }
>   }
> @@ -229,7 +235,7 @@ ipip_input(struct mbuf **mp, int *offp, 
>   ECN_ALLOWED_IPSEC : ECN_ALLOWED;
>   if (!ip_ecn_egress(mode, , >ip_tos)) {
>   DPRINTF(("ipip_input(): ip_ecn_egress() failed"));
> - ipipstat.ipips_pdrops++;
> + ipipstat_inc(ipips_pdrops);

Re: Enable TCP ECN by default (for inbound)

[David Hill]

On Fri, Oct 07, 2016 at 11:05:28PM -0400, David Hill wrote:
> Hello -
> 
> Many operating systems out there are starting to enable TCP ECN by
> default, most inbound-only.  Linux, FreeBSD, Apple MacOS and iOS.
> 
> This diff mimicks the FreeBSD diff found at
> http://marc.info/?l=freebsd-commits-all=146369644516347=2
> 
> It modifies the net.inet.tcp.ecn from an off and on switch to a 
> three-way setting: off, on, and inbound-only (0, 1, 2, respectively).
> 
> It also enables TCP ECN by default for inbound connections that request
> it.
> 
> My home router has TCP ECN enabled. With 36 hours of uptime, netstat -s
> shows:
>   16165 ECN connections accepted
>   1 ECE packet received
>   6 ECE packets sent
>   589 CWR packets received
>   109 CWR packets sent
> 
> This diff is just to see if there is interest in such a change.  If
> people seem to like the idea, perhaps I could update the diff to also:
>   - Remove TCP_ECN kernel option and always compile in support?
>   - Only allow 0, 1, or 2 via the sysctl net.inet.tcp.ecn knob.
>   - Update pf.os
> 
> Thoughts?
>  - David
>  

Here is a diff that doesn't change the default.

Index: netinet/tcp_output.c
===
RCS file: /cvs/src/sys/netinet/tcp_output.c,v
retrieving revision 1.119
diff -u -p -r1.119 tcp_output.c
--- netinet/tcp_output.c9 Feb 2017 15:19:32 -   1.119
+++ netinet/tcp_output.c4 Mar 2017 21:26:02 -
@@ -843,10 +843,12 @@ send:
}
if (!(tp->t_flags & TF_DISABLE_ECN)) {
/*
-* if this is a SYN seg, set ECE and CWR.
-* set only ECE for SYN-ACK if peer supports ECN.
+* If net.inet.tcp.ecn is set to 1 and this is a
+* SYN seg, set both ECE and CWR.
+* Set only ECE for SYN-ACK if peer supports ECN.
 */
-   if ((flags & (TH_SYN|TH_ACK)) == TH_SYN)
+   if (tcp_do_ecn == 1 &&
+   (flags & (TH_SYN|TH_ACK)) == TH_SYN)
flags |= (TH_ECE|TH_CWR);
else if ((tp->t_flags & TF_ECN_PERMIT) &&
 (flags & (TH_SYN|TH_ACK)) == (TH_SYN|TH_ACK))
Index: netinet/tcp_subr.c
===
RCS file: /cvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.160
diff -u -p -r1.160 tcp_subr.c
--- netinet/tcp_subr.c  9 Feb 2017 15:19:32 -   1.160
+++ netinet/tcp_subr.c  4 Mar 2017 21:26:02 -
@@ -110,7 +110,7 @@ int tcp_do_sack = 1;/* RFC 2018 selecti
 #endif
 inttcp_ack_on_push = 0;/* set to enable immediate ACK-on-PUSH */
 #ifdef TCP_ECN
-inttcp_do_ecn = 0; /* RFC3168 ECN enabled/disabled? */
+inttcp_do_ecn = 0; /* RFC3168 TCP ECN 0:off 1:on 2:inbound-only */
 #endif
 inttcp_do_rfc3390 = 2; /* Increase TCP's Initial Window to 10*mss */
 
Index: netinet/tcp_var.h
===
RCS file: /cvs/src/sys/netinet/tcp_var.h,v
retrieving revision 1.122
diff -u -p -r1.122 tcp_var.h
--- netinet/tcp_var.h   9 Feb 2017 15:19:32 -   1.122
+++ netinet/tcp_var.h   4 Mar 2017 21:26:02 -
@@ -703,7 +703,7 @@ extern  int tcp_do_sack;/* SACK enabled/
 extern struct pool sackhl_pool;
 extern int tcp_sackhole_limit; /* max entries for tcp sack queues */
 #endif
-extern int tcp_do_ecn; /* RFC3168 ECN enabled/disabled? */
+extern int tcp_do_ecn; /* RFC3168 ECN: 0:off 1:on 2:inbound-only */
 extern int tcp_do_rfc3390; /* RFC3390 Increasing TCP's Initial Window */
 
 extern struct pool tcpqe_pool;
 

Re: pr_output variable arguments

[David Hill]

Much easier to read as well.  OK

On Fri, Mar 03, 2017 at 03:49:08PM +0100, Alexander Bluhm wrote:
> Hi,
> 
> I prefer compiler checks over variable argument lists.
> So lets convert the pr_output functions.
> 
> ok?
> 
> bluhm
> 
> Index: net/pfkey.c
> ===
> RCS file: /cvs/src/sys/net/pfkey.c,v
> retrieving revision 1.38
> diff -u -p -r1.38 pfkey.c
> --- net/pfkey.c   2 Mar 2017 08:58:24 -   1.38
> +++ net/pfkey.c   3 Mar 2017 14:45:48 -
> @@ -90,10 +90,10 @@ static struct pfkey_version *pfkey_versi
>  
>  struct sockaddr pfkey_addr = { 2, PF_KEY, };
>  
> -/* static struct domain pfkeydomain; */
> -static int pfkey_usrreq(struct socket *socket, int req, struct mbuf *mbuf,
> -struct mbuf *nam, struct mbuf *control, struct proc *);
> -static int pfkey_output(struct mbuf *mbuf, struct socket *socket);
> +int pfkey_usrreq(struct socket *, int , struct mbuf *, struct mbuf *,
> +struct mbuf *, struct proc *);
> +int pfkey_output(struct mbuf *, struct socket *, struct sockaddr *,
> +struct mbuf *);
>  
>  void pfkey_init(void);
>  int pfkey_buildprotosw(void);
> @@ -154,8 +154,9 @@ pfkey_sendup(struct socket *socket, stru
>   return (0);
>  }
>  
> -static int
> -pfkey_output(struct mbuf *mbuf, struct socket *socket)
> +int
> +pfkey_output(struct mbuf *mbuf, struct socket *socket, struct sockaddr 
> *dstaddr,
> +struct mbuf *control)
>  {
>   void *message;
>   int error = 0;
> @@ -230,7 +231,7 @@ pfkey_detach(struct socket *socket, stru
>   return (rval);
>  }
>  
> -static int
> +int
>  pfkey_usrreq(struct socket *socket, int req, struct mbuf *mbuf,
>  struct mbuf *nam, struct mbuf *control, struct proc *p)
>  {
> @@ -266,7 +267,7 @@ static struct protosw pfkey_protosw_temp
>.pr_domain = ,
>.pr_protocol   = -1,
>.pr_flags  = PR_ATOMIC | PR_ADDR,
> -  .pr_output = (void *) pfkey_output,
> +  .pr_output = pfkey_output,
>.pr_usrreq = pfkey_usrreq
>  };
>  
> Index: net/raw_usrreq.c
> ===
> RCS file: /cvs/src/sys/net/raw_usrreq.c,v
> retrieving revision 1.28
> diff -u -p -r1.28 raw_usrreq.c
> --- net/raw_usrreq.c  23 Jan 2017 16:31:24 -  1.28
> +++ net/raw_usrreq.c  3 Mar 2017 14:45:48 -
> @@ -139,7 +139,7 @@ raw_usrreq(struct socket *so, int req, s
>   error = ENOTCONN;
>   break;
>   }
> - error = (*so->so_proto->pr_output)(m, so);
> + error = (*so->so_proto->pr_output)(m, so, NULL, NULL);
>   m = NULL;
>   if (nam)
>   rp->rcb_faddr = 0;
> Index: net/route.h
> ===
> RCS file: /cvs/src/sys/net/route.h,v
> retrieving revision 1.157
> diff -u -p -r1.157 route.h
> --- net/route.h   2 Mar 2017 17:09:21 -   1.157
> +++ net/route.h   3 Mar 2017 14:45:49 -
> @@ -417,7 +417,8 @@ struct sockaddr_in6;
>  struct bfd_config;
>  
>  void  route_init(void);
> -int   route_output(struct mbuf *, ...);
> +int   route_output(struct mbuf *, struct socket *, struct sockaddr *,
> + struct mbuf *);
>  int   route_usrreq(struct socket *, int, struct mbuf *,
>  struct mbuf *, struct mbuf *, struct proc *);
>  void  rt_ifmsg(struct ifnet *);
> Index: net/rtsock.c
> ===
> RCS file: /cvs/src/sys/net/rtsock.c,v
> retrieving revision 1.227
> diff -u -p -r1.227 rtsock.c
> --- net/rtsock.c  3 Mar 2017 14:22:40 -   1.227
> +++ net/rtsock.c  3 Mar 2017 14:45:49 -
> @@ -528,7 +528,8 @@ rt_report(struct rtentry *rt, u_char typ
>  }
>  
>  int
> -route_output(struct mbuf *m, ...)
> +route_output(struct mbuf *m, struct socket *so, struct sockaddr *dstaddr,
> +struct mbuf *control)
>  {
>   struct rt_msghdr*rtm = NULL;
>   struct rtentry  *rt = NULL;
> @@ -536,19 +537,13 @@ route_output(struct mbuf *m, ...)
>   int  plen, len, seq, newgate = 0, error = 0;
>   struct ifnet*ifp = NULL;
>   struct ifaddr   *ifa = NULL;
> - struct socket   *so;
>   struct rawcb*rp = NULL;
>  #ifdef MPLS
>   struct sockaddr_mpls*psa_mpls;
>  #endif
> - va_list  ap;
>   u_inttableid;
>   u_int8_t prio;
>   u_char   vers, type;
> -
> - va_start(ap, m);
> - so = va_arg(ap, struct socket *);
> - va_end(ap);
>  
>   if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
>   (m = m_pullup(m, sizeof(int32_t))) == 0))
> Index: netinet/ip_var.h
> ===
> RCS file: /cvs/src/sys/netinet/ip_var.h,v
> retrieving revision 1.68
> diff -u -p -r1.68 ip_var.h
> --- netinet/ip_var.h  1 

ntp_update_second - removed unused var

[David Hill]

Hello -

time_t *sec is unused in ntp_update_second.
OK?

Index: kern/kern_tc.c
===
RCS file: /cvs/src/sys/kern/kern_tc.c,v
retrieving revision 1.30
diff -u -p -r1.30 kern_tc.c
--- kern/kern_tc.c  9 Feb 2017 20:15:28 -   1.30
+++ kern/kern_tc.c  3 Mar 2017 05:37:17 -
@@ -41,7 +41,7 @@
 
 u_int dummy_get_timecount(struct timecounter *);
 
-void ntp_update_second(int64_t *, time_t *);
+void ntp_update_second(int64_t *);
 int sysctl_tc_hardware(void *, size_t *, void *, size_t);
 int sysctl_tc_choice(void *, size_t *, void *, size_t);
 
@@ -434,7 +434,7 @@ tc_windup(void)
if (i > LARGE_STEP)
i = 2;
for (; i > 0; i--)
-   ntp_update_second(>th_adjustment, );
+   ntp_update_second(>th_adjustment);
 
/* Update the UTC timestamps used by the get*() functions. */
/* XXX shouldn't do this here.  Should force non-`get' versions. */
@@ -617,7 +617,7 @@ sysctl_tc(int *name, u_int namelen, void
 }
 
 void
-ntp_update_second(int64_t *adjust, time_t *sec)
+ntp_update_second(int64_t *adjust)
 {
int64_t adj;
 

Re: C99 initializers for domains

[David Hill]

OK

On Wed, Mar 01, 2017 at 01:40:58PM +0100, Martin Pieuchot wrote:
> I converted the declarations when I audited the existing 'dom_init'
> functions.
> 
> ok?
> 
> Index: kern/uipc_proto.c
> ===
> RCS file: /cvs/src/sys/kern/uipc_proto.c,v
> retrieving revision 1.12
> diff -u -p -r1.12 uipc_proto.c
> --- kern/uipc_proto.c 22 Feb 2017 19:34:42 -  1.12
> +++ kern/uipc_proto.c 1 Mar 2017 12:05:35 -
> @@ -37,9 +37,9 @@
>  #include 
>  #include 
>  #include 
> -#include  
> +#include 
>  #include 
> -
> +
>  #include 
>  
>  /*
> @@ -72,6 +72,11 @@ struct protosw unixsw[] = {
>  }
>  };
>  
> -struct domain unixdomain =
> -{ AF_LOCAL, "unix", 0, unp_externalize, unp_dispose,
> -  unixsw, [nitems(unixsw)] };
> +struct domain unixdomain = {
> +  .dom_family = AF_LOCAL,
> +  .dom_name = "unix",
> +  .dom_externalize = unp_externalize,
> +  .dom_dispose = unp_dispose,
> +  .dom_protosw = unixsw,
> +  .dom_protoswNPROTOSW = [nitems(unixsw)]
> +};
> Index: net/pfkey.c
> ===
> RCS file: /cvs/src/sys/net/pfkey.c,v
> retrieving revision 1.37
> diff -u -p -r1.37 pfkey.c
> --- net/pfkey.c   22 Feb 2017 19:34:42 -  1.37
> +++ net/pfkey.c   1 Mar 2017 12:11:31 -
> @@ -256,13 +256,9 @@ pfkey_usrreq(struct socket *socket, int 
>  }
>  
>  struct domain pfkeydomain = {
> - PF_KEY,
> - "PF_KEY",
> - pfkey_init, /* init */
> - NULL, /* externalize */
> - NULL, /* dispose */
> - NULL, /* protosw */
> - NULL, /* protoswNPROTOSW */
> +  .dom_family = PF_KEY,
> +  .dom_name = "PF_KEY",
> +  .dom_init = pfkey_init,
>  };
>  
>  static struct protosw pfkey_protosw_template = {
> Index: net/rtsock.c
> ===
> RCS file: /cvs/src/sys/net/rtsock.c,v
> retrieving revision 1.223
> diff -u -p -r1.223 rtsock.c
> --- net/rtsock.c  22 Feb 2017 19:34:42 -  1.223
> +++ net/rtsock.c  1 Mar 2017 12:33:46 -
> @@ -1666,6 +1666,10 @@ struct protosw routesw[] = {
>  }
>  };
>  
> -struct domain routedomain =
> -{ PF_ROUTE, "route", route_init, 0, 0,
> -  routesw, [nitems(routesw)] };
> +struct domain routedomain = {
> +  .dom_family = PF_ROUTE,
> +  .dom_name = "route",
> +  .dom_init = route_init,
> +  .dom_protosw = routesw,
> +  .dom_protoswNPROTOSW = [nitems(routesw)]
> +};
> Index: netinet/in_proto.c
> ===
> RCS file: /cvs/src/sys/netinet/in_proto.c,v
> retrieving revision 1.73
> diff -u -p -r1.73 in_proto.c
> --- netinet/in_proto.c22 Feb 2017 19:34:42 -  1.73
> +++ netinet/in_proto.c1 Mar 2017 12:07:53 -
> @@ -438,8 +438,12 @@ struct protosw inetsw[] = {
>  }
>  };
>  
> -struct domain inetdomain =
> -{ AF_INET, "internet", 0, 0, 0,
> -  inetsw, [nitems(inetsw)],
> -  sizeof(struct sockaddr_in),
> -  offsetof(struct sockaddr_in, sin_addr), 32 };
> +struct domain inetdomain = {
> +  .dom_family = AF_INET,
> +  .dom_name = "internet",
> +  .dom_protosw = inetsw,
> +  .dom_protoswNPROTOSW = [nitems(inetsw)],
> +  .dom_rtkeylen = sizeof(struct sockaddr_in),
> +  .dom_rtoffset = offsetof(struct sockaddr_in, sin_addr),
> +  .dom_maxplen = 32
> +};
> Index: netinet6/in6_proto.c
> ===
> RCS file: /cvs/src/sys/netinet6/in6_proto.c,v
> retrieving revision 1.89
> diff -u -p -r1.89 in6_proto.c
> --- netinet6/in6_proto.c  22 Feb 2017 19:34:42 -  1.89
> +++ netinet6/in6_proto.c  1 Mar 2017 12:10:23 -
> @@ -340,13 +340,17 @@ struct protosw inet6sw[] = {
>  }
>  };
>  
> -struct domain inet6domain =
> -{ AF_INET6, "internet6", 0, 0, 0,
> -  (struct protosw *)inet6sw,
> -  (struct protosw *)[nitems(inet6sw)],
> -  sizeof(struct sockaddr_in6),
> -  offsetof(struct sockaddr_in6, sin6_addr), 128,
> -  in6_domifattach, in6_domifdetach, };
> +struct domain inet6domain = {
> +  .dom_family = AF_INET6,
> +  .dom_name = "internet6",
> +  .dom_protosw = (struct protosw *)inet6sw,
> +  .dom_protoswNPROTOSW = (struct protosw *)[nitems(inet6sw)],
> +  .dom_rtkeylen = sizeof(struct sockaddr_in6),
> +  .dom_rtoffset = offsetof(struct sockaddr_in6, sin6_addr),
> +  .dom_maxplen = 128,
> +  .dom_ifattach = in6_domifattach,
> +  .dom_ifdetach = in6_domifdetach
> +};
>  
>  /*
>   * Internet configuration info
> Index: netmpls/mpls_proto.c
> ===
> RCS file: /cvs/src/sys/netmpls/mpls_proto.c,v
> retrieving revision 1.15
> diff -u -p -r1.15 mpls_proto.c
> --- netmpls/mpls_proto.c  27 Feb 2017 19:16:56 -  1.15
> +++ netmpls/mpls_proto.c  1 Mar 2017 12:14:17 -
> @@ -46,9 +46,9 @@
>   * MPLS address family: needed for the routing table
>   */
>  struct domain mplsdomain = {
> - 

Re: make sosetopt responsible for m_free

[David Hill]

On Mon, Feb 06, 2017 at 01:16:45PM +0100, Martin Pieuchot wrote:
> On 03/02/17(Fri) 11:02, David Hill wrote:
> > On Fri, Feb 03, 2017 at 09:50:40AM +0100, Martin Pieuchot wrote:
> > > On 02/02/17(Thu) 12:12, David Hill wrote:
> > > > On Thu, Feb 02, 2017 at 09:34:07AM +0100, Martin Pieuchot wrote:
> > > > > On 01/02/17(Wed) 19:27, David Hill wrote:
> > > > > > Hello -
> > > > > > 
> > > > > > This diff makes sosetopt responsible for m_free which is much 
> > > > > > simpler.
> > > > > > Requested by bluhm@ 
> > > > > 
> > > > > I'd suggest to move the m_free(9) calls to sys_setsockopt().  This
> > > > > simplifies the existing code even more and will make it easier to use
> > > > > the stack for this temporary storage.
> > > > > 
> > > > 
> > > > New diff with mpi@'s suggestion. 
> > > 
> > > You forgot NFS and BFD that should now call m_free(9) after sosetopt(9). 
> > >
> > 
> > Indeed!  Now with BFD and NFS...
> 
> You're introducing a use after-free in ip_pcbopts().  You need to
> allocate/copy the mbuf there.
> 
> I must say I'm a bit afraid of this change because the amount of code it
> touches.  There might be another use after free somewhere that I missed.
> 
> Maybe we should first split our huge *ctloutput functions.  
> 
> One easy move is to split setopt/getopt.
> 
> Then introduce more per-protocol functions instead of having everything
> in ip{,6}_ctloutput().
> 
> For example move all the IPSEC craziness out of these functions.  Same
> thing with ICMP6...  This might sound superfluous but it will help
> if/when we decide to have a fine graining for different subsystems.
> 
> I'd also suggest to change the 'struct protosw' declaration to use C99
> initializer.  So we can have:
> 
>   .pr_ctloutput = ipsec_ctloutput
> 
> This would allow us to grep for "pr_ctloutput" (or pr_setopt) and know
> directly which functions to review.
>

If you are OK with first splitting each *ctloutput into *getopt/*setopt,
I will send each diff individually to make review easier.

Here is the first split, route_ctloutput.

Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.222
diff -u -p -r1.222 rtsock.c
--- net/rtsock.c1 Feb 2017 20:59:47 -   1.222
+++ net/rtsock.c21 Feb 2017 15:52:07 -
@@ -98,6 +98,8 @@ struct walkarg {
caddr_t w_where, w_tmem;
 };
 
+introute_getopt(struct socket *, int, int, struct mbuf *);
+introute_setopt(struct socket *, int, int, struct mbuf *);
 introute_ctloutput(int, struct socket *, int, int, struct mbuf *);
 void   route_input(struct mbuf *m0, sa_family_t);
 introute_arp_conflict(struct rtentry *, struct rt_addrinfo *);
@@ -233,62 +235,86 @@ route_usrreq(struct socket *so, int req,
 }
 
 int
-route_ctloutput(int op, struct socket *so, int level, int optname,
-struct mbuf *m)
+route_getopt(struct socket *so, int level, int optname, struct mbuf *m)
+{
+   struct routecb *rop = sotoroutecb(so);
+   int error = 0;
+
+   if (level != AF_ROUTE)
+   return EINVAL;
+
+   switch (optname) {
+   case ROUTE_MSGFILTER:
+   m->m_len = sizeof(unsigned int);
+   *mtod(m, unsigned int *) = rop->msgfilter;
+   break;
+   case ROUTE_TABLEFILTER:
+   m->m_len = sizeof(unsigned int);
+   *mtod(m, unsigned int *) = rop->rtableid;
+   break;
+   default:
+   error = ENOPROTOOPT;
+   break;
+   }
+   return error;
+}
+
+int
+route_setopt(struct socket *so, int level, int optname, struct mbuf *m)
 {
struct routecb *rop = sotoroutecb(so);
int error = 0;
unsigned int tid;
 
if (level != AF_ROUTE) {
-   error = EINVAL;
-   if (op == PRCO_SETOPT && m)
-   m_free(m);
-   return (error);
+   m_free(m);
+   return EINVAL;
}
 
-   switch (op) {
-   case PRCO_SETOPT:
-   switch (optname) {
-   case ROUTE_MSGFILTER:
-   if (m == NULL || m->m_len != sizeof(unsigned int))
-   error = EINVAL;
-   else
-   rop->msgfilter = *mtod(m, unsigned int *);
-   break;
-   case ROUTE_TABLEFILTER:
-   if (m == NULL || m->m_len != sizeof(unsigned int)) {
-   error = EINVAL;
-   break;
-   

Re: c99 initialize struct protosw

[David Hill]

Here is an updated diff without explicitly setting 0/NULL.

Index: kern/uipc_proto.c
===
RCS file: /cvs/src/sys/kern/uipc_proto.c,v
retrieving revision 1.11
diff -u -p -r1.11 uipc_proto.c
--- kern/uipc_proto.c   5 Feb 2017 07:57:08 -   1.11
+++ kern/uipc_proto.c   21 Feb 2017 00:42:46 -
@@ -49,20 +49,26 @@
 extern struct domain unixdomain;   /* or at least forward */
 
 struct protosw unixsw[] = {
-{ SOCK_STREAM, ,PF_LOCAL,   
PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
-  0,   0,  0,  0,
-  uipc_usrreq,
-  0,   0,  0,  0,
+{
+  .pr_type = SOCK_STREAM,
+  .pr_domain   = ,
+  .pr_protocol  = PF_LOCAL,
+  .pr_flags= PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
+  .pr_usrreq   = uipc_usrreq,
 },
-{ SOCK_SEQPACKET,,  PF_LOCAL,   
PR_ATOMIC|PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
-  0,   0,  0,  0,
-  uipc_usrreq,
-  0,   0,  0,  0,
+{
+  .pr_type = SOCK_SEQPACKET,
+  .pr_domain   = ,
+  .pr_protocol = PF_LOCAL,
+  .pr_flags= PR_ATOMIC|PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
+  .pr_usrreq   = uipc_usrreq,
 },
-{ SOCK_DGRAM,  ,PF_LOCAL,   PR_ATOMIC|PR_ADDR|PR_RIGHTS,
-  0,   0,  0,  0,
-  uipc_usrreq,
-  0,   0,  0,  0,
+{
+  .pr_type = SOCK_DGRAM,
+  .pr_domain   = ,
+  .pr_protocol = PF_LOCAL,
+  .pr_flags= PR_ATOMIC|PR_ADDR|PR_RIGHTS,
+  .pr_usrreq   = uipc_usrreq,
 }
 };
 
Index: net/pfkey.c
===
RCS file: /cvs/src/sys/net/pfkey.c,v
retrieving revision 1.36
diff -u -p -r1.36 pfkey.c
--- net/pfkey.c 24 Jan 2017 10:08:30 -  1.36
+++ net/pfkey.c 21 Feb 2017 00:42:46 -
@@ -266,20 +266,12 @@ struct domain pfkeydomain = {
 };
 
 static struct protosw pfkey_protosw_template = {
-   SOCK_RAW,
-   ,
-   -1, /* protocol */
-   PR_ATOMIC | PR_ADDR,
-   NULL, /* input */
-   (void *) pfkey_output,
-   NULL, /* ctlinput */
-   NULL, /* ctloutput */
-   pfkey_usrreq,
-   NULL, /* init */
-   NULL, /* fasttimo */
-   NULL, /* slowtimo */
-   NULL, /* drain */
-   NULL/* sysctl */
+  .pr_type = SOCK_RAW,
+  .pr_domain   = ,
+  .pr_protocol = -1,
+  .pr_flags= PR_ATOMIC | PR_ADDR,
+  .pr_output   = (void *) pfkey_output,
+  .pr_usrreq   = pfkey_usrreq
 };
 
 int
Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.222
diff -u -p -r1.222 rtsock.c
--- net/rtsock.c1 Feb 2017 20:59:47 -   1.222
+++ net/rtsock.c21 Feb 2017 00:42:47 -
@@ -1654,11 +1654,15 @@ sysctl_rtable_rtstat(void *oldp, size_t 
 extern struct domain routedomain;  /* or at least forward */
 
 struct protosw routesw[] = {
-{ SOCK_RAW,,   0,  PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
-  0,   route_output,   0,  route_ctloutput,
-  route_usrreq,
-  raw_init,0,  0,  0,
-  sysctl_rtable,
+{
+  .pr_type = SOCK_RAW,
+  .pr_domain   = ,
+  .pr_flags= PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
+  .pr_output   = route_output,
+  .pr_ctloutput= route_ctloutput,
+  .pr_usrreq   = route_usrreq,
+  .pr_init = raw_init,
+  .pr_sysctl   = sysctl_rtable
 }
 };
 
Index: netinet/in_proto.c
===
RCS file: /cvs/src/sys/netinet/in_proto.c,v
retrieving revision 1.72
diff -u -p -r1.72 in_proto.c
--- netinet/in_proto.c  29 Jan 2017 19:58:47 -  1.72
+++ netinet/in_proto.c  21 Feb 2017 00:42:47 -
@@ -175,138 +175,267 @@
 u_char ip_protox[IPPROTO_MAX];
 
 struct protosw inetsw[] = {
-{ 0,   ,0,  0,
-  0,   0,  0,  0,
-  0,
-  ip_init, 0,  ip_slowtimo,ip_drain,   ip_sysctl
-},
-{ SOCK_DGRAM,  ,IPPROTO_UDP,PR_ATOMIC|PR_ADDR|PR_SPLICE,
-  udp_input,   0,  udp_ctlinput,   ip_ctloutput,
-  udp_usrreq,
-  udp_init,0,  0,  0,  udp_sysctl
-},
-{ SOCK_STREAM, ,IPPROTO_TCP,
PR_CONNREQUIRED|PR_WANTRCVD|PR_ABRTACPTDIS|PR_SPLICE,
-  tcp_input,   0,  tcp_ctlinput,   tcp_ctloutput,
-  tcp_usrreq,
-  tcp_init,0,  tcp_slowtimo,   0,  tcp_sysctl
-},
-{ SOCK_RAW,,IPPROTO_RAW,PR_ATOMIC|PR_ADDR,
-  rip_input,   rip_output, 0,  rip_ctloutput,
-  rip_usrreq,
-  0,   0,  0,  0,
-},
-{ SOCK_RAW,,IPPROTO_ICMP,   PR_ATOMIC|PR_ADDR,
-  icmp_input,  rip_output, 0,  rip_ctloutput,
-  rip_usrreq,
-  icmp_init,   0,  0,  0,  icmp_sysctl
+{
+  .pr_domain   = ,
+  .pr_init = ip_init,
+  .pr_slowtimo = 

ip6_ctloutput - check invalid level first

[David Hill]

Hello -

The *ctloutput functions all check for an invalid level first, except
ip6_ctloutput.  This diff makes ip6_ctloutput do the same.   As a bonus,
we save on indentation.  No functional change.

Index: netinet6/ip6_output.c
===
RCS file: /cvs/src/sys/netinet6/ip6_output.c,v
retrieving revision 1.225
diff -u -p -r1.225 ip6_output.c
--- netinet6/ip6_output.c   9 Feb 2017 15:19:32 -   1.225
+++ netinet6/ip6_output.c   20 Feb 2017 21:02:47 -
@@ -1062,64 +1062,69 @@ ip6_ctloutput(int op, struct socket *so,
privileged = (inp->inp_socket->so_state & SS_PRIV);
uproto = (int)so->so_proto->pr_protocol;
 
-   if (level == IPPROTO_IPV6) {
-   switch (op) {
-   case PRCO_SETOPT:
+   if (level != IPPROTO_IPV6) {
+   if (op == PRCO_SETOPT)
+   m_free(m);
+   return (EINVAL);
+   }
+
+   switch (op) {
+   case PRCO_SETOPT:
+   switch (optname) {
+   /*
+* Use of some Hop-by-Hop options or some
+* Destination options, might require special
+* privilege.  That is, normal applications
+* (without special privilege) might be forbidden
+* from setting certain options in outgoing packets,
+* and might never see certain options in received
+* packets. [RFC 2292 Section 6]
+* KAME specific note:
+*  KAME prevents non-privileged users from sending or
+*  receiving ANY hbh/dst options in order to avoid
+*  overhead of parsing options in the kernel.
+*/
+   case IPV6_RECVHOPOPTS:
+   case IPV6_RECVDSTOPTS:
+   if (!privileged) {
+   error = EPERM;
+   break;
+   }
+   /* FALLTHROUGH */
+   case IPV6_UNICAST_HOPS:
+   case IPV6_MINHOPCOUNT:
+   case IPV6_HOPLIMIT:
+
+   case IPV6_RECVPKTINFO:
+   case IPV6_RECVHOPLIMIT:
+   case IPV6_RECVRTHDR:
+   case IPV6_RECVPATHMTU:
+   case IPV6_RECVTCLASS:
+   case IPV6_V6ONLY:
+   case IPV6_AUTOFLOWLABEL:
+   case IPV6_RECVDSTPORT:
+   if (m == NULL || m->m_len != sizeof(int)) {
+   error = EINVAL;
+   break;
+   }
+   optval = *mtod(m, int *);
switch (optname) {
-   /*
-* Use of some Hop-by-Hop options or some
-* Destination options, might require special
-* privilege.  That is, normal applications
-* (without special privilege) might be forbidden
-* from setting certain options in outgoing packets,
-* and might never see certain options in received
-* packets. [RFC 2292 Section 6]
-* KAME specific note:
-*  KAME prevents non-privileged users from sending or
-*  receiving ANY hbh/dst options in order to avoid
-*  overhead of parsing options in the kernel.
-*/
-   case IPV6_RECVHOPOPTS:
-   case IPV6_RECVDSTOPTS:
-   if (!privileged) {
-   error = EPERM;
-   break;
-   }
-   /* FALLTHROUGH */
-   case IPV6_UNICAST_HOPS:
-   case IPV6_MINHOPCOUNT:
-   case IPV6_HOPLIMIT:
 
-   case IPV6_RECVPKTINFO:
-   case IPV6_RECVHOPLIMIT:
-   case IPV6_RECVRTHDR:
-   case IPV6_RECVPATHMTU:
-   case IPV6_RECVTCLASS:
-   case IPV6_V6ONLY:
-   case IPV6_AUTOFLOWLABEL:
-   case IPV6_RECVDSTPORT:
-   if (m == NULL || m->m_len != sizeof(int)) {
+   case IPV6_UNICAST_HOPS:
+   if (optval < -1 || optval >= 256)
error = EINVAL;
-   break;
+   else {
+   /* -1 = kernel default */
+   inp->inp_hops = optval;
}
-   optval = *mtod(m, int *);
-   switch (optname) {
-
-  

Re: c99 initialize struct protosw

[David Hill]

On Sat, Feb 18, 2017 at 09:22:44PM -0500, David Hill wrote:
> Hello -
> 
> This moves the 'struct protosw' declarations to use C99 initializers.
> Requested by mpi@
>

Apologies for not explaining the reason for the diff.

It would allow us to easily grep for a member, such as pr_ctloutput,
and know which functions need review.  The goal is to introduce more
per-protocol functions instead of having everything in
ip{,6}_ctloutput(), with the next diff splitting setopt and getopt up.

- David

c99 initialize struct protosw

[David Hill]

Hello -

This moves the 'struct protosw' declarations to use C99 initializers.
Requested by mpi@

Index: kern/uipc_proto.c
===
RCS file: /cvs/src/sys/kern/uipc_proto.c,v
retrieving revision 1.11
diff -u -p -r1.11 uipc_proto.c
--- kern/uipc_proto.c   5 Feb 2017 07:57:08 -   1.11
+++ kern/uipc_proto.c   19 Feb 2017 02:15:00 -
@@ -49,20 +49,53 @@
 extern struct domain unixdomain;   /* or at least forward */
 
 struct protosw unixsw[] = {
-{ SOCK_STREAM, ,PF_LOCAL,   
PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
-  0,   0,  0,  0,
-  uipc_usrreq,
-  0,   0,  0,  0,
+{
+  .pr_type = SOCK_STREAM,
+  .pr_domain   = ,
+  .pr_protocol  = PF_LOCAL,
+  .pr_flags= PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
+  .pr_input= NULL,
+  .pr_output   = NULL,
+  .pr_ctlinput = NULL,
+  .pr_ctloutput= NULL,
+  .pr_usrreq   = uipc_usrreq,
+  .pr_init = NULL,
+  .pr_fasttimo = NULL,
+  .pr_slowtimo = NULL,
+  .pr_drain= NULL,
+  .pr_sysctl   = NULL
 },
-{ SOCK_SEQPACKET,,  PF_LOCAL,   
PR_ATOMIC|PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
-  0,   0,  0,  0,
-  uipc_usrreq,
-  0,   0,  0,  0,
+{
+  .pr_type = SOCK_SEQPACKET,
+  .pr_domain   = ,
+  .pr_protocol = PF_LOCAL,
+  .pr_flags= PR_ATOMIC|PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS,
+  .pr_input= NULL,
+  .pr_output   = NULL,
+  .pr_ctlinput = NULL,
+  .pr_ctloutput= NULL,
+  .pr_usrreq   = uipc_usrreq,
+  .pr_init = NULL,
+  .pr_fasttimo = NULL,
+  .pr_slowtimo = NULL,
+  .pr_drain= NULL,
+  .pr_sysctl   = NULL
 },
-{ SOCK_DGRAM,  ,PF_LOCAL,   PR_ATOMIC|PR_ADDR|PR_RIGHTS,
-  0,   0,  0,  0,
-  uipc_usrreq,
-  0,   0,  0,  0,
+{
+  .pr_type = SOCK_DGRAM,
+  .pr_domain   = ,
+  .pr_protocol = PF_LOCAL,
+  .pr_flags= PR_ATOMIC|PR_ADDR|PR_RIGHTS,
+  .pr_input= NULL,
+  .pr_output   = NULL,
+  .pr_ctlinput = NULL,
+  .pr_ctloutput= NULL,
+  .pr_usrreq   = uipc_usrreq,
+  .pr_init = NULL,
+  .pr_fasttimo = NULL,
+  .pr_slowtimo = NULL,
+  .pr_drain= NULL,
+  .pr_sysctl   = NULL
 }
 };
 
Index: net/pfkey.c
===
RCS file: /cvs/src/sys/net/pfkey.c,v
retrieving revision 1.36
diff -u -p -r1.36 pfkey.c
--- net/pfkey.c 24 Jan 2017 10:08:30 -  1.36
+++ net/pfkey.c 19 Feb 2017 02:15:00 -
@@ -266,20 +266,20 @@ struct domain pfkeydomain = {
 };
 
 static struct protosw pfkey_protosw_template = {
-   SOCK_RAW,
-   ,
-   -1, /* protocol */
-   PR_ATOMIC | PR_ADDR,
-   NULL, /* input */
-   (void *) pfkey_output,
-   NULL, /* ctlinput */
-   NULL, /* ctloutput */
-   pfkey_usrreq,
-   NULL, /* init */
-   NULL, /* fasttimo */
-   NULL, /* slowtimo */
-   NULL, /* drain */
-   NULL/* sysctl */
+  .pr_type = SOCK_RAW,
+  .pr_domain   = ,
+  .pr_protocol = -1,
+  .pr_flags= PR_ATOMIC | PR_ADDR,
+  .pr_input= NULL,
+  .pr_output   = (void *) pfkey_output,
+  .pr_ctlinput = NULL,
+  .pr_ctloutput= NULL,
+  .pr_usrreq   = pfkey_usrreq,
+  .pr_init = NULL,
+  .pr_fasttimo = NULL,
+  .pr_slowtimo = NULL,
+  .pr_drain= NULL,
+  .pr_sysctl   = NULL
 };
 
 int
Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.222
diff -u -p -r1.222 rtsock.c
--- net/rtsock.c1 Feb 2017 20:59:47 -   1.222
+++ net/rtsock.c19 Feb 2017 02:15:00 -
@@ -1654,13 +1654,23 @@ sysctl_rtable_rtstat(void *oldp, size_t 
 extern struct domain routedomain;  /* or at least forward */
 
 struct protosw routesw[] = {
-{ SOCK_RAW,,   0,  PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
-  0,   route_output,   0,  route_ctloutput,
-  route_usrreq,
-  raw_init,0,  0,  0,
-  sysctl_rtable,
+{
+  .pr_type = SOCK_RAW,
+  .pr_domain   = ,
+  .pr_protocol = 0,
+  .pr_flags= PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
+  .pr_input= NULL,
+  .pr_output   = route_output,
+  .pr_ctlinput = NULL,
+  .pr_ctloutput= route_ctloutput,
+  .pr_usrreq   = route_usrreq,
+  .pr_init = raw_init,
+  .pr_fasttimo = NULL,
+  .pr_slowtimo = NULL,
+  .pr_drain= NULL,
+  .pr_sysctl   = sysctl_rtable
 }
-};
+}; 
 
 struct domain routedomain =
 { PF_ROUTE, "route", route_init, 0, 0,
Index: netinet/in_proto.c
===
RCS file: /cvs/src/sys/netinet/in_proto.c,v
retrieving revision 1.72
diff -u -p -r1.72 in_proto.c
--- netinet/in_proto.c  29 Jan 2017 19:58:47 -  1.72
+++ netinet/in_proto.c  19 Feb 2017 02:15:00 -
@@ -175,138 +175,380 @@
 u_char ip_protox[IPPROTO_MAX];
 
 struct 

Re: ipcomp ipv6 output

[David Hill]

OK.

On Fri, Feb 17, 2017 at 03:09:25PM +0100, Alexander Bluhm wrote:
> Hi,
> 
> When using ipcomp(4) with IPv6 transport mode I get this error:
> 
> ping6: sendmsg: Protocol family not supported
> 
> The break in the switch statement is missing since rev 1.1 from 2001.
> So I am the first one testing this.
> 
> ok?
> 
> bluhm
> 
> Index: netinet/ip_ipcomp.c
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/ip_ipcomp.c,v
> retrieving revision 1.54
> diff -u -p -r1.54 ip_ipcomp.c
> --- netinet/ip_ipcomp.c   9 Feb 2017 00:43:58 -   1.54
> +++ netinet/ip_ipcomp.c   17 Feb 2017 13:44:11 -
> @@ -411,6 +411,7 @@ ipcomp_output(struct mbuf *m, struct tdb
>   ipcompstat.ipcomps_toobig++;
>   return EMSGSIZE;
>   }
> + break;
>  #endif /* INET6 */
>  
>   default:
> 

Re: ipsec output failure counter

[David Hill]

OK once the XXX from ip_esp.c is removed too.


On Tue, Feb 07, 2017 at 06:15:03PM +0100, Alexander Bluhm wrote:
> Hi,
> 
> As mentioned before, IPsec packets could be dropped unaccounted if
> output after crypto failed.  Add a counter for that case.
> 
> ok?
> 
> bluhm
> 
> Index: sys/netinet/ip_ah.c
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/ip_ah.c,v
> retrieving revision 1.126
> diff -u -p -r1.126 ip_ah.c
> --- sys/netinet/ip_ah.c   7 Feb 2017 15:10:48 -   1.126
> +++ sys/netinet/ip_ah.c   7 Feb 2017 16:55:11 -
> @@ -1247,8 +1247,8 @@ ah_output_cb(struct cryptop *crp)
>   /* No longer needed. */
>   crypto_freereq(crp);
>  
> - ipsp_process_done(m, tdb);
> - /* XXX missing error counter if ipsp_process_done() drops packet */
> + if (ipsp_process_done(m, tdb))
> + ahstat.ahs_outfail++;
>   NET_UNLOCK(s);
>  
>   baddone:
> Index: sys/netinet/ip_ah.h
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/ip_ah.h,v
> retrieving revision 1.33
> diff -u -p -r1.33 ip_ah.h
> --- sys/netinet/ip_ah.h   10 Jan 2010 12:43:07 -  1.33
> +++ sys/netinet/ip_ah.h   7 Feb 2017 16:55:11 -
> @@ -38,8 +38,7 @@
>  #ifndef _NETINET_IP_AH_H_
>  #define _NETINET_IP_AH_H_
>  
> -struct ahstat
> -{
> +struct ahstat {
>  u_int32_tahs_hdrops; /* Packet shorter than header shows */
>  u_int32_tahs_nopf;   /* Protocol family not supported */
>  u_int32_tahs_notdb;
> @@ -58,10 +57,10 @@ struct ahstat
>  u_int32_tahs_toobig; /* Packet got larger than IP_MAXPACKET 
> */
>  u_int32_tahs_pdrops; /* Packet blocked due to policy */
>  u_int32_tahs_crypto; /* Crypto processing failure */
> +u_int32_tahs_outfail;/* Packet output failure */
>  };
>  
> -struct ah
> -{
> +struct ah {
>  u_int8_t   ah_nh;
>  u_int8_t   ah_hl;
>  u_int16_t  ah_rv;
> Index: sys/netinet/ip_esp.c
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/ip_esp.c,v
> retrieving revision 1.144
> diff -u -p -r1.144 ip_esp.c
> --- sys/netinet/ip_esp.c  7 Feb 2017 15:10:48 -   1.144
> +++ sys/netinet/ip_esp.c  7 Feb 2017 16:55:11 -
> @@ -1088,7 +1088,8 @@ esp_output_cb(struct cryptop *crp)
>   crypto_freereq(crp);
>  
>   /* Call the IPsec input callback. */
> - ipsp_process_done(m, tdb);
> + if (ipsp_process_done(m, tdb))
> + espstat.esps_outfail++;
>   /* XXX missing error counter if ipsp_process_done() drops packet */
>   NET_UNLOCK(s);
>   return;
> Index: sys/netinet/ip_esp.h
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/ip_esp.h,v
> retrieving revision 1.43
> diff -u -p -r1.43 ip_esp.h
> --- sys/netinet/ip_esp.h  2 Sep 2016 09:39:32 -   1.43
> +++ sys/netinet/ip_esp.h  7 Feb 2017 16:55:11 -
> @@ -38,8 +38,7 @@
>  #ifndef _NETINET_IP_ESP_H_
>  #define _NETINET_IP_ESP_H_
>  
> -struct espstat
> -{
> +struct espstat {
>  u_int32_tesps_hdrops;/* Packet shorter than header shows */
>  u_int32_tesps_nopf;  /* Protocol family not supported */
>  u_int32_tesps_notdb;
> @@ -63,6 +62,7 @@ struct espstat
>  u_int32_tesps_udpencout; /* Output ESP-in-UDP packets */
>  u_int32_tesps_udpinval;  /* Invalid input ESP-in-UDP packets */
>  u_int32_tesps_udpneeded; /* Trying to use a ESP-in-UDP TDB */
> +u_int32_tesps_outfail;   /* Packet output failure */
>  };
>  
>  /*
> Index: sys/netinet/ip_ipcomp.c
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/ip_ipcomp.c,v
> retrieving revision 1.51
> diff -u -p -r1.51 ip_ipcomp.c
> --- sys/netinet/ip_ipcomp.c   7 Feb 2017 15:10:48 -   1.51
> +++ sys/netinet/ip_ipcomp.c   7 Feb 2017 16:55:11 -
> @@ -579,8 +579,8 @@ ipcomp_output_cb(struct cryptop *crp)
>   if (rlen < crp->crp_olen) {
>   /* Compression was useless, we have lost time. */
>   crypto_freereq(crp);
> - ipsp_process_done(m, tdb);
> - /* XXX missing counter if ipsp_process_done() drops packet */
> + if (ipsp_process_done(m, tdb))
> + ipcompstat.ipcomps_outfail++;
>   NET_UNLOCK(s);
>   return;
>   }
> @@ -628,8 +628,8 @@ ipcomp_output_cb(struct cryptop *crp)
>   /* Release the crypto descriptor. */
>   crypto_freereq(crp);
>  
> - ipsp_process_done(m, tdb);
> - /* XXX missing error counter if ipsp_process_done() drops packet */
> + if (ipsp_process_done(m, tdb))
> + ipcompstat.ipcomps_outfail++;
>   

Re: make sosetopt responsible for m_free

[David Hill]

On Fri, Feb 03, 2017 at 09:50:40AM +0100, Martin Pieuchot wrote:
> On 02/02/17(Thu) 12:12, David Hill wrote:
> > On Thu, Feb 02, 2017 at 09:34:07AM +0100, Martin Pieuchot wrote:
> > > On 01/02/17(Wed) 19:27, David Hill wrote:
> > > > Hello -
> > > > 
> > > > This diff makes sosetopt responsible for m_free which is much simpler.
> > > > Requested by bluhm@ 
> > > 
> > > I'd suggest to move the m_free(9) calls to sys_setsockopt().  This
> > > simplifies the existing code even more and will make it easier to use
> > > the stack for this temporary storage.
> > > 
> > 
> > New diff with mpi@'s suggestion. 
> 
> You forgot NFS and BFD that should now call m_free(9) after sosetopt(9). 
>

Indeed!  Now with BFD and NFS...

Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.176
diff -u -p -r1.176 uipc_socket.c
--- kern/uipc_socket.c  1 Feb 2017 20:59:47 -   1.176
+++ kern/uipc_socket.c  3 Feb 2017 16:00:26 -
@@ -1551,16 +1551,15 @@ sowwakeup(struct socket *so)
 }
 
 int
-sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
+sosetopt(struct socket *so, int level, int optname, struct mbuf *m)
 {
int s, error = 0;
-   struct mbuf *m = m0;
 
if (level != SOL_SOCKET) {
if (so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, m0);
+   level, optname, m);
NET_UNLOCK(s);
return (error);
}
@@ -1707,7 +1706,7 @@ sosetopt(struct socket *so, int level, i
level = dom->dom_protosw->pr_protocol;
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_SETOPT, so, level, optname, m0);
+   (PRCO_SETOPT, so, level, optname, m);
NET_UNLOCK(s);
return (error);
}
@@ -1739,14 +1738,11 @@ sosetopt(struct socket *so, int level, i
if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
(*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, m0);
+   level, optname, m);
NET_UNLOCK(s);
-   m = NULL;   /* freed by protocol */
}
}
 bad:
-   if (m)
-   (void) m_free(m);
return (error);
 }
 
Index: kern/uipc_syscalls.c
===
RCS file: /cvs/src/sys/kern/uipc_syscalls.c,v
retrieving revision 1.148
diff -u -p -r1.148 uipc_syscalls.c
--- kern/uipc_syscalls.c26 Jan 2017 01:58:00 -  1.148
+++ kern/uipc_syscalls.c3 Feb 2017 16:00:26 -
@@ -962,19 +962,13 @@ sys_setsockopt(struct proc *p, void *v, 
goto bad;
}
}
-   if (m == NULL) {
-   error = ENOBUFS;
-   goto bad;
-   }
error = copyin(SCARG(uap, val), mtod(m, caddr_t),
SCARG(uap, valsize));
-   if (error) {
+   if (error)
goto bad;
-   }
m->m_len = SCARG(uap, valsize);
}
error = sosetopt(fp->f_data, SCARG(uap, level), SCARG(uap, name), m);
-   m = NULL;
 bad:
m_freem(m);
FRELE(fp, p);
Index: net/bfd.c
===
RCS file: /cvs/src/sys/net/bfd.c,v
retrieving revision 1.58
diff -u -p -r1.58 bfd.c
--- net/bfd.c   24 Jan 2017 10:08:30 -  1.58
+++ net/bfd.c   3 Feb 2017 16:00:26 -
@@ -418,7 +418,7 @@ bfd_listener(struct bfd_config *bfd, uns
struct sockaddr_in  *sin;
struct sockaddr_in6 *sin6;
struct socket   *so;
-   struct mbuf *m = NULL, *mopt = NULL;
+   struct mbuf *m = NULL, *mopt;
int *ip, error;
 
/* sa_family and sa_len must be equal */
@@ -437,6 +437,7 @@ bfd_listener(struct bfd_config *bfd, uns
ip = mtod(mopt, int *);
*ip = MAXTTL;
error = sosetopt(so, IPPROTO_IP, IP_MINTTL, mopt);
+   m_free(mopt);
if (error) {
printf("%s: sosetopt error %d\n",
__func__, error);
@@ -487,7 +488,7 @@ b

Re: make sosetopt responsible for m_free

[David Hill]

On Thu, Feb 02, 2017 at 09:34:07AM +0100, Martin Pieuchot wrote:
> On 01/02/17(Wed) 19:27, David Hill wrote:
> > Hello -
> > 
> > This diff makes sosetopt responsible for m_free which is much simpler.
> > Requested by bluhm@ 
> 
> I'd suggest to move the m_free(9) calls to sys_setsockopt().  This
> simplifies the existing code even more and will make it easier to use
> the stack for this temporary storage.
> 

New diff with mpi@'s suggestion. 

Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.176
diff -u -p -r1.176 uipc_socket.c
--- kern/uipc_socket.c  1 Feb 2017 20:59:47 -   1.176
+++ kern/uipc_socket.c  2 Feb 2017 16:59:45 -
@@ -1551,16 +1551,15 @@ sowwakeup(struct socket *so)
 }
 
 int
-sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
+sosetopt(struct socket *so, int level, int optname, struct mbuf *m)
 {
int s, error = 0;
-   struct mbuf *m = m0;
 
if (level != SOL_SOCKET) {
if (so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, m0);
+   level, optname, m);
NET_UNLOCK(s);
return (error);
}
@@ -1707,7 +1706,7 @@ sosetopt(struct socket *so, int level, i
level = dom->dom_protosw->pr_protocol;
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_SETOPT, so, level, optname, m0);
+   (PRCO_SETOPT, so, level, optname, m);
NET_UNLOCK(s);
return (error);
}
@@ -1739,14 +1738,11 @@ sosetopt(struct socket *so, int level, i
if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
(*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, m0);
+   level, optname, m);
NET_UNLOCK(s);
-   m = NULL;   /* freed by protocol */
}
}
 bad:
-   if (m)
-   (void) m_free(m);
return (error);
 }
 
Index: kern/uipc_syscalls.c
===
RCS file: /cvs/src/sys/kern/uipc_syscalls.c,v
retrieving revision 1.148
diff -u -p -r1.148 uipc_syscalls.c
--- kern/uipc_syscalls.c26 Jan 2017 01:58:00 -  1.148
+++ kern/uipc_syscalls.c2 Feb 2017 16:59:45 -
@@ -962,19 +962,13 @@ sys_setsockopt(struct proc *p, void *v, 
goto bad;
}
}
-   if (m == NULL) {
-   error = ENOBUFS;
-   goto bad;
-   }
error = copyin(SCARG(uap, val), mtod(m, caddr_t),
SCARG(uap, valsize));
-   if (error) {
+   if (error)
goto bad;
-   }
m->m_len = SCARG(uap, valsize);
}
error = sosetopt(fp->f_data, SCARG(uap, level), SCARG(uap, name), m);
-   m = NULL;
 bad:
m_freem(m);
FRELE(fp, p);
Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.222
diff -u -p -r1.222 rtsock.c
--- net/rtsock.c1 Feb 2017 20:59:47 -   1.222
+++ net/rtsock.c2 Feb 2017 16:59:45 -
@@ -240,12 +240,8 @@ route_ctloutput(int op, struct socket *s
int error = 0;
unsigned int tid;
 
-   if (level != AF_ROUTE) {
-   error = EINVAL;
-   if (op == PRCO_SETOPT && m)
-   m_free(m);
-   return (error);
-   }
+   if (level != AF_ROUTE)
+   return EINVAL;
 
switch (op) {
case PRCO_SETOPT:
@@ -271,7 +267,6 @@ route_ctloutput(int op, struct socket *s
error = ENOPROTOOPT;
break;
}
-   m_free(m);
break;
case PRCO_GETOPT:
switch (optname) {
Index: netinet/ip_mroute.c
===
RCS file: /cvs/src/sys/netinet/ip_mroute.c,v
retrieving revision 1.108
diff -u -p -r1.108 ip_mroute.c
--- netinet/ip_mroute.c 1 Feb 2017 20:59:47 -   1.108
+++ netinet/ip_mroute.c 2 Feb 2017 16:59:46 -
@@ -209,7 +209,6 @@ ip_mrouter_set(struct socket *so

make sosetopt responsible for m_free

[David Hill]

Hello -

This diff makes sosetopt responsible for m_free which is much simpler.
Requested by bluhm@ 

Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.176
diff -u -p -r1.176 uipc_socket.c
--- kern/uipc_socket.c  1 Feb 2017 20:59:47 -   1.176
+++ kern/uipc_socket.c  2 Feb 2017 00:13:23 -
@@ -1562,6 +1562,7 @@ sosetopt(struct socket *so, int level, i
error = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
level, optname, m0);
NET_UNLOCK(s);
+   m_free(m0);
return (error);
}
error = ENOPROTOOPT;
@@ -1709,6 +1710,7 @@ sosetopt(struct socket *so, int level, i
error = (*so->so_proto->pr_ctloutput)
(PRCO_SETOPT, so, level, optname, m0);
NET_UNLOCK(s);
+   m_free(m0);
return (error);
}
error = ENOPROTOOPT;
@@ -1741,7 +1743,8 @@ sosetopt(struct socket *so, int level, i
(*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
level, optname, m0);
NET_UNLOCK(s);
-   m = NULL;   /* freed by protocol */
+   m_free(m0);
+   m = NULL;
}
}
 bad:
Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.222
diff -u -p -r1.222 rtsock.c
--- net/rtsock.c1 Feb 2017 20:59:47 -   1.222
+++ net/rtsock.c2 Feb 2017 00:13:23 -
@@ -240,12 +240,8 @@ route_ctloutput(int op, struct socket *s
int error = 0;
unsigned int tid;
 
-   if (level != AF_ROUTE) {
-   error = EINVAL;
-   if (op == PRCO_SETOPT && m)
-   m_free(m);
-   return (error);
-   }
+   if (level != AF_ROUTE)
+   return EINVAL;
 
switch (op) {
case PRCO_SETOPT:
@@ -271,7 +267,6 @@ route_ctloutput(int op, struct socket *s
error = ENOPROTOOPT;
break;
}
-   m_free(m);
break;
case PRCO_GETOPT:
switch (optname) {
Index: netinet/ip_mroute.c
===
RCS file: /cvs/src/sys/netinet/ip_mroute.c,v
retrieving revision 1.108
diff -u -p -r1.108 ip_mroute.c
--- netinet/ip_mroute.c 1 Feb 2017 20:59:47 -   1.108
+++ netinet/ip_mroute.c 2 Feb 2017 00:13:23 -
@@ -209,7 +209,6 @@ ip_mrouter_set(struct socket *so, int op
break;
}
 
-   m_free(m);
return (error);
 }
 
Index: netinet/ip_output.c
===
RCS file: /cvs/src/sys/netinet/ip_output.c,v
retrieving revision 1.335
diff -u -p -r1.335 ip_output.c
--- netinet/ip_output.c 1 Feb 2017 20:59:47 -   1.335
+++ netinet/ip_output.c 2 Feb 2017 00:13:23 -
@@ -853,11 +853,10 @@ ip_ctloutput(int op, struct socket *so, 
int error = 0;
u_int rtid = 0;
 
-   if (level != IPPROTO_IP) {
-   error = EINVAL;
-   if (op == PRCO_SETOPT)
-   (void) m_free(m);
-   } else switch (op) {
+   if (level != IPPROTO_IP)
+   return EINVAL;
+   
+   switch (op) {
case PRCO_SETOPT:
switch (optname) {
case IP_OPTIONS:
@@ -1073,7 +1072,6 @@ ip_ctloutput(int op, struct socket *so, 
error = ENOPROTOOPT;
break;
}
-   m_free(m);
break;
 
case PRCO_GETOPT:
@@ -1235,12 +1233,11 @@ ip_pcbopts(struct mbuf **pcbopt, struct 
 
/* turn off any old options */
m_free(*pcbopt);
-   *pcbopt = 0;
+   *pcbopt = NULL;
if (m == NULL || m->m_len == 0) {
/*
 * Only turning off any previous options.
 */
-   m_free(m);
return (0);
}
 
@@ -1316,7 +1313,6 @@ ip_pcbopts(struct mbuf **pcbopt, struct 
return (0);
 
 bad:
-   (void)m_free(m);
return (EINVAL);
 }
 
Index: netinet/raw_ip.c
===
RCS file: /cvs/src/sys/netinet/raw_ip.c,v
retrieving revision 1.95
diff -u -p -r1.95 raw_ip.c
--- netinet/raw_ip.c1 Feb 2017 20:59:47 -   1.95
+++ netinet/raw_ip.c2 Feb 2017 00:13:23 -
@@ -305,11 +305,8 @@ rip_ctloutput(int op, struct socket *so,
int error = 0;
int dir;
 
-   if (level != 

Re: Help with the NET_LOCK()

[David Hill]

On Tue, Jan 31, 2017 at 09:11:37PM +0100, Alexander Bluhm wrote:
> On Tue, Jan 31, 2017 at 12:14:35PM -0500, David Hill wrote:
> > with mpi@'s suggestion to pass a struct mbuf * 
> 
> We call mbuf variables m and mbuf pointer mp.  So you should rename
> *mp to m.
> 
> The different policy who has to free the mbuf with
> if (op == PRCO_SETOPT)
> m_free(*mp);
> is not nice.  I think it would be better if all the freeing is
> done in sosetopt and sogetopt.  But this requires more thought
> and should not be in this diff.  A possible next step.
> 
> bluhm
> 

I was thinking sosetopt in a separate diff..

Updated diff.

Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.175
diff -u -p -r1.175 uipc_socket.c
--- kern/uipc_socket.c  27 Jan 2017 20:31:42 -  1.175
+++ kern/uipc_socket.c  31 Jan 2017 20:29:37 -
@@ -1560,7 +1560,7 @@ sosetopt(struct socket *so, int level, i
if (so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, );
+   level, optname, m0);
NET_UNLOCK(s);
return (error);
}
@@ -1707,7 +1707,7 @@ sosetopt(struct socket *so, int level, i
level = dom->dom_protosw->pr_protocol;
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_SETOPT, so, level, optname, );
+   (PRCO_SETOPT, so, level, optname, m0);
NET_UNLOCK(s);
return (error);
}
@@ -1739,7 +1739,7 @@ sosetopt(struct socket *so, int level, i
if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
(*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, );
+   level, optname, m0);
NET_UNLOCK(s);
m = NULL;   /* freed by protocol */
}
@@ -1758,11 +1758,19 @@ sogetopt(struct socket *so, int level, i
 
if (level != SOL_SOCKET) {
if (so->so_proto && so->so_proto->pr_ctloutput) {
+   m = m_get(M_WAIT, MT_SOOPTS);
+   m->m_len = 0;
+
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_GETOPT, so,
-   level, optname, mp);
+   level, optname, m);
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   m_free(m);
+   return (error);
+   }
+   *mp = m;
+   return (0);
} else
return (ENOPROTOOPT);
} else {
@@ -1835,7 +1843,6 @@ sogetopt(struct socket *so, int level, i
}
 
case SO_RTABLE:
-   (void)m_free(m);
if (so->so_proto && so->so_proto->pr_domain &&
so->so_proto->pr_domain->dom_protosw &&
so->so_proto->pr_ctloutput) {
@@ -1844,12 +1851,16 @@ sogetopt(struct socket *so, int level, i
level = dom->dom_protosw->pr_protocol;
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_GETOPT, so, level, optname, mp);
+   (PRCO_GETOPT, so, level, optname, m);
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   (void)m_free(m);
+   return (error);
+   }
+   break;
}
+   (void)m_free(m);
return (ENOPROTOOPT);
-   break;
 
 #ifdef SOCKET_SPLICE
case SO_SPLICE:
@@ -1880,7 +1891,6 @@ sogetopt(struct socket *so, int level, i
}
(void)m_free(m);
return (EOPNOTSUPP);
-   break;
 
default:

Re: Help with the NET_LOCK()

[David Hill]

On Tue, Jan 31, 2017 at 10:43:26AM +0100, Martin Pieuchot wrote:
> On 27/01/17(Fri) 14:33, David Hill wrote:
> > [...] 
> > Forgot a file...   Try this:
> 
> Is it now possible to pass a 'struct mbuf *' instead of a 'struct mbuf **'
> to the pr_ctloutput() functions?
> 
> Changing the signature would ensure we do not miss a call.  This would
> also simplify the SETOPT case.
>

with mpi@'s suggestion to pass a struct mbuf * 

Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.175
diff -u -p -r1.175 uipc_socket.c
--- kern/uipc_socket.c  27 Jan 2017 20:31:42 -  1.175
+++ kern/uipc_socket.c  31 Jan 2017 17:09:42 -
@@ -1560,7 +1560,7 @@ sosetopt(struct socket *so, int level, i
if (so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, );
+   level, optname, m0);
NET_UNLOCK(s);
return (error);
}
@@ -1707,7 +1707,7 @@ sosetopt(struct socket *so, int level, i
level = dom->dom_protosw->pr_protocol;
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_SETOPT, so, level, optname, );
+   (PRCO_SETOPT, so, level, optname, m0);
NET_UNLOCK(s);
return (error);
}
@@ -1739,7 +1739,7 @@ sosetopt(struct socket *so, int level, i
if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
NET_LOCK(s);
(*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so,
-   level, optname, );
+   level, optname, m0);
NET_UNLOCK(s);
m = NULL;   /* freed by protocol */
}
@@ -1758,11 +1758,19 @@ sogetopt(struct socket *so, int level, i
 
if (level != SOL_SOCKET) {
if (so->so_proto && so->so_proto->pr_ctloutput) {
+   m = m_get(M_WAIT, MT_SOOPTS);
+   m->m_len = 0;
+
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_GETOPT, so,
-   level, optname, mp);
+   level, optname, m);
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   m_free(m);
+   return (error);
+   }
+   *mp = m;
+   return (0);
} else
return (ENOPROTOOPT);
} else {
@@ -1835,7 +1843,6 @@ sogetopt(struct socket *so, int level, i
}
 
case SO_RTABLE:
-   (void)m_free(m);
if (so->so_proto && so->so_proto->pr_domain &&
so->so_proto->pr_domain->dom_protosw &&
so->so_proto->pr_ctloutput) {
@@ -1844,12 +1851,16 @@ sogetopt(struct socket *so, int level, i
level = dom->dom_protosw->pr_protocol;
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_GETOPT, so, level, optname, mp);
+   (PRCO_GETOPT, so, level, optname, m);
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   (void)m_free(m);
+   return (error);
+   }
+   break;
}
+   (void)m_free(m);
return (ENOPROTOOPT);
-   break;
 
 #ifdef SOCKET_SPLICE
case SO_SPLICE:
@@ -1880,7 +1891,6 @@ sogetopt(struct socket *so, int level, i
}
(void)m_free(m);
return (EOPNOTSUPP);
-   break;
 
default:
(void)m_free(m);
Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.221
diff -u -p -r1.221 rtsock.c
--- net/rtsock.c31 Jan 2017 10:24

uvm: free sizes diff

[David Hill]

Hello -

The following diff adds free sizes to free() calls in uvm/.  Only one
remaining in uvm/.

Index: uvm/uvm_amap.c
===
RCS file: /cvs/src/sys/uvm/uvm_amap.c,v
retrieving revision 1.78
diff -u -p -r1.78 uvm_amap.c
--- uvm/uvm_amap.c  8 Oct 2016 16:19:44 -   1.78
+++ uvm/uvm_amap.c  30 Jan 2017 21:17:22 -
@@ -368,7 +368,7 @@ amap_alloc1(int slots, int waitf, int la
return(amap);
 
 fail1:
-   free(amap->am_buckets, M_UVMAMAP, 0);
+   free(amap->am_buckets, M_UVMAMAP, buckets * sizeof(*amap->am_buckets));
TAILQ_FOREACH_SAFE(chunk, >am_chunks, ac_list, tmp)
pool_put(_amap_chunk_pool, chunk);
pool_put(_amap_pool, amap);
@@ -414,7 +414,7 @@ amap_free(struct vm_amap *amap)
 
 #ifdef UVM_AMAP_PPREF
if (amap->am_ppref && amap->am_ppref != PPREF_NONE)
-   free(amap->am_ppref, M_UVMAMAP, 0);
+   free(amap->am_ppref, M_UVMAMAP, amap->am_nslot * sizeof(int));
 #endif
 
if (UVM_AMAP_SMALL(amap))
Index: uvm/uvm_aobj.c
===
RCS file: /cvs/src/sys/uvm/uvm_aobj.c,v
retrieving revision 1.84
diff -u -p -r1.84 uvm_aobj.c
--- uvm/uvm_aobj.c  24 Sep 2016 18:40:29 -  1.84
+++ uvm/uvm_aobj.c  30 Jan 2017 21:17:22 -
@@ -403,7 +403,7 @@ uao_free(struct uvm_aobj *aobj)
uvmexp.swpgonly--;
}
}
-   free(aobj->u_swslots, M_UVMAOBJ, 0);
+   free(aobj->u_swslots, M_UVMAOBJ, aobj->u_pages * sizeof(int));
}
 
/* finally free the aobj itself */
@@ -532,7 +532,7 @@ uao_shrink_array(struct uvm_object *uobj
for (i = 0; i < pages; i++)
new_swslots[i] = aobj->u_swslots[i];
 
-   free(aobj->u_swslots, M_UVMAOBJ, 0);
+   free(aobj->u_swslots, M_UVMAOBJ, aobj->u_pages * sizeof(int));
 
aobj->u_swslots = new_swslots;
aobj->u_pages = pages;
@@ -585,7 +585,7 @@ uao_grow_array(struct uvm_object *uobj, 
for (i = 0; i < aobj->u_pages; i++)
new_swslots[i] = aobj->u_swslots[i];
 
-   free(aobj->u_swslots, M_UVMAOBJ, 0);
+   free(aobj->u_swslots, M_UVMAOBJ, aobj->u_pages * sizeof(int));
 
aobj->u_swslots = new_swslots;
aobj->u_pages = pages;
@@ -664,7 +664,7 @@ uao_grow_convert(struct uvm_object *uobj
}
}
 
-   free(old_swslots, M_UVMAOBJ, 0);
+   free(old_swslots, M_UVMAOBJ, aobj->u_pages * sizeof(int));
aobj->u_pages = pages;
 
return 0;

Re: Help with the NET_LOCK()

[David Hill]

On Fri, Jan 27, 2017 at 08:09:36PM +0100, Hrvoje Popovski wrote:
> On 27.1.2017. 19:14, David Hill wrote:
> >> splassert: yield: want 0 have 1
> >> Starting stack trace...
> >> yield() at yield+0xac
> >> pool_get() at pool_get+0x1ca
> >> m_get() at m_get+0x28
> >> ip_ctloutput() at ip_ctloutput+0x4bf
> >> sogetopt() at sogetopt+0x7e
> >> sys_getsockopt() at sys_getsockopt+0xbf
> >> syscall() at syscall+0x27b
> >> --- syscall (number 118) ---
> >> end of kernel
> >> end trace frame: 0x3, count: 250
> >> 0x978bdd844a:
> >> End of stack trace.
> >>  
> >>
> > Attempted to solve this and am running with this diff:
> 
> 
> Hi,
> 
> i applied you patch and i'm still seeing this trace
> 
> 
> splassert: yield: want 0 have 1
> Starting stack trace...
> yield() at yield+0xac
> pool_get() at pool_get+0x1ca
> m_get() at m_get+0x28
> ip_ctloutput() at ip_ctloutput+0x4bf
> sogetopt() at sogetopt+0xa1
> sys_getsockopt() at sys_getsockopt+0xbf
> syscall() at syscall+0x27b
> --- syscall (number 118) ---
> end of kernel
> end trace frame: 0x3, count: 250
> 0x178f12db8f1a:
> End of stack trace.
> 
> 
> and this one i'm seeing for first time, maybe because of this diff
> 
> splassert: yield: want 0 have 1
> Starting stack trace...
> yield() at yield+0xac
> malloc() at malloc+0x406
> ip_setmoptions() at ip_setmoptions+0x248
> ip_ctloutput() at ip_ctloutput+0x461
> sosetopt() at sosetopt+0x8e
> sys_setsockopt() at sys_setsockopt+0x12d
> syscall() at syscall+0x27b
> --- syscall (number 105) ---
> end of kernel
> end trace frame: 0x1f83, count: 250
> 0x91243a37f1a:
> End of stack trace.
>

Forgot a file...   Try this:

 
Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.174
diff -u -p -r1.174 uipc_socket.c
--- kern/uipc_socket.c  26 Jan 2017 00:08:50 -  1.174
+++ kern/uipc_socket.c  27 Jan 2017 19:30:31 -
@@ -1758,11 +1758,19 @@ sogetopt(struct socket *so, int level, i
 
if (level != SOL_SOCKET) {
if (so->so_proto && so->so_proto->pr_ctloutput) {
+   m = m_get(M_WAIT, MT_SOOPTS);
+   m->m_len = 0;
+
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_GETOPT, so,
-   level, optname, mp);
+   level, optname, );
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   m_free(m);
+   return (error);
+   }
+   *mp = m;
+   return (0);
} else
return (ENOPROTOOPT);
} else {
@@ -1835,21 +1843,25 @@ sogetopt(struct socket *so, int level, i
}
 
case SO_RTABLE:
-   (void)m_free(m);
if (so->so_proto && so->so_proto->pr_domain &&
so->so_proto->pr_domain->dom_protosw &&
so->so_proto->pr_ctloutput) {
struct domain *dom = so->so_proto->pr_domain;
 
level = dom->dom_protosw->pr_protocol;
+   
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_GETOPT, so, level, optname, mp);
+   (PRCO_GETOPT, so, level, optname, );
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   (void)m_free(m);
+   return (error);
+   }
+   break;
}
+   (void)m_free(m);
return (ENOPROTOOPT);
-   break;
 
 #ifdef SOCKET_SPLICE
case SO_SPLICE:
@@ -1880,7 +1892,6 @@ sogetopt(struct socket *so, int level, i
}
(void)m_free(m);
return (EOPNOTSUPP);
-   break;
 
default:
(void)m_free(m);
Index: net/rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.220
diff -u -p -r1.220 rtsock.c
--- net/rtsock.c24 Jan 2017 00:17:14 -  1

Re: Help with the NET_LOCK()

[David Hill]

On Wed, Jan 25, 2017 at 11:14:57AM -0500, David Hill wrote:
> On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> > I just enabled the NET_LOCK() again and I'm looking for test reports.
> > Please go build a kernel from sources or wait for the next snapshot,
> > run it and report back.
> > 
> > If you're looking for some small coding tasks related to the NET_LOCK()
> > just do:
> > 
> > # sysctl kern.splassert=2
> > # sysctl kern.pool_debug=2
> > 
> > Then watch for the traces on your console.
> > 
> > You'll see something like:
> > 
> > Starting stack trace...
> > yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
> > yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
> > pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
> > m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
> > doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
> > sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
> > syscall() at syscall+0x250
> > 
> > This means accept(2) is doing a memory allocation that can sleep, here
> > with m_get(9), while holding the NET_LOCK().  Even if these should be
> > ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> > be allocated beforehand or simply use the stack for that.
> > 
> > Cheers,
> > Martin
> >
> 
> splassert: yield: want 0 have 1
> Starting stack trace...
> yield() at yield+0xac
> pool_get() at pool_get+0x1ca
> m_get() at m_get+0x28
> ip_ctloutput() at ip_ctloutput+0x4bf
> sogetopt() at sogetopt+0x7e
> sys_getsockopt() at sys_getsockopt+0xbf
> syscall() at syscall+0x27b
> --- syscall (number 118) ---
> end of kernel
> end trace frame: 0x3, count: 250
> 0x978bdd844a:
> End of stack trace.
>  
>

Attempted to solve this and am running with this diff:

Index: kern/uipc_socket.c
===
RCS file: /cvs/src/sys/kern/uipc_socket.c,v
retrieving revision 1.174
diff -u -p -r1.174 uipc_socket.c
--- kern/uipc_socket.c  26 Jan 2017 00:08:50 -  1.174
+++ kern/uipc_socket.c  27 Jan 2017 18:08:26 -
@@ -1758,11 +1758,19 @@ sogetopt(struct socket *so, int level, i
 
if (level != SOL_SOCKET) {
if (so->so_proto && so->so_proto->pr_ctloutput) {
+   m = m_get(M_WAIT, MT_SOOPTS);
+   m->m_len = 0;
+
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)(PRCO_GETOPT, so,
-   level, optname, mp);
+   level, optname, );
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   m_free(m);
+   return (error);
+   }
+   *mp = m;
+   return (0);
} else
return (ENOPROTOOPT);
} else {
@@ -1835,21 +1843,25 @@ sogetopt(struct socket *so, int level, i
}
 
case SO_RTABLE:
-   (void)m_free(m);
if (so->so_proto && so->so_proto->pr_domain &&
so->so_proto->pr_domain->dom_protosw &&
so->so_proto->pr_ctloutput) {
struct domain *dom = so->so_proto->pr_domain;
 
level = dom->dom_protosw->pr_protocol;
+   
NET_LOCK(s);
error = (*so->so_proto->pr_ctloutput)
-   (PRCO_GETOPT, so, level, optname, mp);
+   (PRCO_GETOPT, so, level, optname, );
NET_UNLOCK(s);
-   return (error);
+   if (error) {
+   (void)m_free(m);
+   return (error);
+   }
+   break;
}
+   (void)m_free(m);
return (ENOPROTOOPT);
-   break;
 
 #ifdef SOCKET_SPLICE
case SO_SPLICE:
@@ -1880,7 +1892,6 @@ sogetopt(struct socket *so, int level, i
}
(void)m_free(m);
return (EOPNOTSUPP);
-   break;
 
default:
(void)m_free(m);
Index: net/rtsock.c

ip6_output: copypktopts - drop canwait

[David Hill]

Hi -

copypktopts is only called by ip6_setpktopts with M_NOWAIT.  Drop
canwait to make this simpler and more readable.

Index: ip6_output.c
===
RCS file: /cvs/src/sys/netinet6/ip6_output.c,v
retrieving revision 1.221
diff -u -p -r1.221 ip6_output.c
--- ip6_output.c19 Jan 2017 14:49:19 -  1.221
+++ ip6_output.c26 Jan 2017 20:18:00 -
@@ -127,7 +127,7 @@ int ip6_insertfraghdr(struct mbuf *, str
 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *);
-int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
+int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *);
 static __inline u_int16_t __attribute__((__unused__))
 in6_cksum_phdr(const struct in6_addr *, const struct in6_addr *,
 u_int32_t, u_int32_t);
@@ -1839,22 +1839,22 @@ ip6_clearpktopts(struct ip6_pktopts *pkt
 do {\
if (src->type) {\
size_t hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 
3;\
-   dst->type = malloc(hlen, M_IP6OPT, canwait);\
-   if (dst->type == NULL && canwait == M_NOWAIT)\
+   dst->type = malloc(hlen, M_IP6OPT, M_NOWAIT);\
+   if (dst->type == NULL)\
goto bad;\
memcpy(dst->type, src->type, hlen);\
}\
 } while (/*CONSTCOND*/ 0)
 
 int
-copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
+copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src)
 {
dst->ip6po_hlim = src->ip6po_hlim;
dst->ip6po_tclass = src->ip6po_tclass;
dst->ip6po_flags = src->ip6po_flags;
if (src->ip6po_pktinfo) {
dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
-   M_IP6OPT, canwait);
+   M_IP6OPT, M_NOWAIT);
if (dst->ip6po_pktinfo == NULL)
goto bad;
*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
@@ -2256,7 +2256,7 @@ ip6_setpktopts(struct mbuf *control, str
 * but we can allow this since this option should be rarely
 * used.
 */
-   if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
+   if ((error = copypktopts(opt, stickyopt)) != 0)
return (error);
}
 

Re: Help with the NET_LOCK()

[David Hill]

On Wed, Jan 25, 2017 at 11:11:03AM -0500, David Hill wrote:
> On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> > I just enabled the NET_LOCK() again and I'm looking for test reports.
> > Please go build a kernel from sources or wait for the next snapshot,
> > run it and report back.
> > 
> > If you're looking for some small coding tasks related to the NET_LOCK()
> > just do:
> > 
> > # sysctl kern.splassert=2
> > # sysctl kern.pool_debug=2
> > 
> > Then watch for the traces on your console.
> > 
> > You'll see something like:
> > 
> > Starting stack trace...
> > yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
> > yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
> > pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
> > m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
> > doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
> > sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
> > syscall() at syscall+0x250
> > 
> > This means accept(2) is doing a memory allocation that can sleep, here
> > with m_get(9), while holding the NET_LOCK().  Even if these should be
> > ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> > be allocated beforehand or simply use the stack for that.
> > 
> > Cheers,
> > Martin
> >
> 
> splassert: yield: want 0 have 1
> Starting stack trace...
> yield() at yield+0xac
> malloc() at malloc+0x406
> ip6_setmoptions() at ip6_setmoptions+0x65
> ip6_ctloutput() at ip6_ctloutput+0x6d9
> sosetopt() at sosetopt+0x8e
> sys_setsockopt() at sys_setsockopt+0x12d
> syscall() at syscall+0x27b
> --- syscall (number 105) ---
> end of kernel
> end trace frame: 0x1b1987be0028, count: 250
> 0x1b197d24b80a:
> End of stack trace.
>  
>

Switch to NOWAIT?  Check is already there...
 
Index: netinet6/ip6_output.c
===
RCS file: /cvs/src/sys/netinet6/ip6_output.c,v
retrieving revision 1.221
diff -u -p -r1.221 ip6_output.c
--- netinet6/ip6_output.c   19 Jan 2017 14:49:19 -  1.221
+++ netinet6/ip6_output.c   26 Jan 2017 01:10:58 -
@@ -1902,7 +1902,7 @@ ip6_setmoptions(int optname, struct ip6_
 * allocate one and initialize to default values.
 */
im6o = (struct ip6_moptions *)
-   malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
+   malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
 
if (im6o == NULL)
return (ENOBUFS);

Re: Help with the NET_LOCK()

[David Hill]

On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> I just enabled the NET_LOCK() again and I'm looking for test reports.
> Please go build a kernel from sources or wait for the next snapshot,
> run it and report back.
> 
> If you're looking for some small coding tasks related to the NET_LOCK()
> just do:
> 
>   # sysctl kern.splassert=2
>   # sysctl kern.pool_debug=2
>   
> Then watch for the traces on your console.
> 
> You'll see something like:
> 
>   Starting stack trace...
>   yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
>   yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
>   pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
>   m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
>   doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
>   sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
>   syscall() at syscall+0x250
>   
> This means accept(2) is doing a memory allocation that can sleep, here
> with m_get(9), while holding the NET_LOCK().  Even if these should be
> ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> be allocated beforehand or simply use the stack for that.
> 
> Cheers,
> Martin
>

Allocate the mbuf beforehand.  Also, move the setting of nflag closer
to where its value is used.

Index: uipc_syscalls.c
===
RCS file: /cvs/src/sys/kern/uipc_syscalls.c,v
retrieving revision 1.147
diff -u -p -r1.147 uipc_syscalls.c
--- uipc_syscalls.c 25 Jan 2017 07:35:31 -  1.147
+++ uipc_syscalls.c 26 Jan 2017 01:03:13 -
@@ -265,7 +265,7 @@ doaccept(struct proc *p, int sock, struc
 {
struct filedesc *fdp = p->p_fd;
struct file *fp, *headfp;
-   struct mbuf *nam = NULL;
+   struct mbuf *nam;
socklen_t namelen;
int error, s, tmpfd;
struct socket *head, *so;
@@ -288,7 +288,8 @@ doaccept(struct proc *p, int sock, struc
return (error);
}
 
-redo:
+   nam = m_get(M_WAIT, MT_SONAME);
+   
NET_LOCK(s);
head = headfp->f_data;
if (isdnssocket(head) || (head->so_options & SO_ACCEPTCONN) == 0) {
@@ -318,30 +319,16 @@ redo:
goto out;
}
 
-   /* Figure out whether the new socket should be non-blocking. */
-   nflag = flags & SOCK_NONBLOCK_INHERIT ? (headfp->f_flag & FNONBLOCK)
-   : (flags & SOCK_NONBLOCK ? FNONBLOCK : 0);
-
-   nam = m_get(M_WAIT, MT_SONAME);
-
-   /*
-* Check whether the queue emptied while we slept: m_get() may have
-* blocked, allowing the connection to be reset or another thread or
-* process to accept it.  If so, start over.
-*/
-   if (head->so_qlen == 0) {
-   NET_UNLOCK(s);
-   m_freem(nam);
-   nam = NULL;
-   goto redo;
-   }
-
/*
 * Do not sleep after we have taken the socket out of the queue.
 */
so = TAILQ_FIRST(>so_q);
if (soqremque(so, 1) == 0)
panic("accept");
+
+   /* Figure out whether the new socket should be non-blocking. */
+   nflag = flags & SOCK_NONBLOCK_INHERIT ? (headfp->f_flag & FNONBLOCK)
+   : (flags & SOCK_NONBLOCK ? FNONBLOCK : 0);
 
/* connection has been removed from the listen queue */
KNOTE(>so_rcv.sb_sel.si_note, 0);

Re: Help with the NET_LOCK()

[David Hill]

On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> I just enabled the NET_LOCK() again and I'm looking for test reports.
> Please go build a kernel from sources or wait for the next snapshot,
> run it and report back.
> 
> If you're looking for some small coding tasks related to the NET_LOCK()
> just do:
> 
>   # sysctl kern.splassert=2
>   # sysctl kern.pool_debug=2
>   
> Then watch for the traces on your console.
> 
> You'll see something like:
> 
>   Starting stack trace...
>   yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
>   yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
>   pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
>   m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
>   doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
>   sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
>   syscall() at syscall+0x250
>   
> This means accept(2) is doing a memory allocation that can sleep, here
> with m_get(9), while holding the NET_LOCK().  Even if these should be
> ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> be allocated beforehand or simply use the stack for that.
> 
> Cheers,
> Martin
>

splassert: yield: want 0 have 1
Starting stack trace...
yield() at yield+0xac
malloc() at malloc+0x406
import_identity() at import_identity+0x30
import_identities() at import_identities+0x62
pfkeyv2_send() at pfkeyv2_send+0x1074
pfkey_output() at pfkey_output+0x87
raw_usrreq() at raw_usrreq+0x232
sosend() at sosend+0x2ec
dofilewritev() at dofilewritev+0x205
sys_writev() at sys_writev+0x6d
syscall() at syscall+0x27b
--- syscall (number 121) ---
end of kernel
end trace frame: 0x1ef4e34e5300, count: 246
0x1ef227c2f3aa:
End of stack trace.
 

Re: Help with the NET_LOCK() - socreate

[David Hill]

On Wed, Jan 25, 2017 at 07:04:19PM +0100, Alexander Bluhm wrote:
> On Wed, Jan 25, 2017 at 11:02:22AM -0500, David Hill wrote:
> > splassert: yield: want 0 have 1
> > Starting stack trace...
> > yield() at yield+0xac
> > pool_get() at pool_get+0x1ca
> > socreate() at socreate+0xba
> > sys_socket() at sys_socket+0x135
> > syscall() at syscall+0x27b
> > --- syscall (number 97) ---
> > end of kernel
> > end trace frame: 0x1b1a16c05800, count: 252
> > 0x1b197d23277a:
> > End of stack trace.
> 
> This one looks easy.  We do not need a lock to setup the still
> private so structure.
> 
> ok?
> 
> bluhm
> 

Which uncovers: 

splassert: yield: want 0 have 1
Starting stack trace...
yield() at yield+0xac
malloc() at malloc+0x406
route_usrreq() at route_usrreq+0xb0
socreate() at socreate+0x15f
sys_socket() at sys_socket+0x135
syscall() at syscall+0x27b
--- syscall (number 97) ---
end of kernel
end trace frame: 0x1b171444a720, count: 251
0x1b19e6cfd5ca:
End of stack trace.
 

Re: Help with the NET_LOCK()

[David Hill]

On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> I just enabled the NET_LOCK() again and I'm looking for test reports.
> Please go build a kernel from sources or wait for the next snapshot,
> run it and report back.
> 
> If you're looking for some small coding tasks related to the NET_LOCK()
> just do:
> 
>   # sysctl kern.splassert=2
>   # sysctl kern.pool_debug=2
>   
> Then watch for the traces on your console.
> 
> You'll see something like:
> 
>   Starting stack trace...
>   yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
>   yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
>   pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
>   m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
>   doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
>   sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
>   syscall() at syscall+0x250
>   
> This means accept(2) is doing a memory allocation that can sleep, here
> with m_get(9), while holding the NET_LOCK().  Even if these should be
> ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> be allocated beforehand or simply use the stack for that.
> 
> Cheers,
> Martin
>

splassert: yield: want 0 have 1
Starting stack trace...
yield() at yield+0xac
pool_get() at pool_get+0x1ca
m_get() at m_get+0x28
ip_ctloutput() at ip_ctloutput+0x4bf
sogetopt() at sogetopt+0x7e
sys_getsockopt() at sys_getsockopt+0xbf
syscall() at syscall+0x27b
--- syscall (number 118) ---
end of kernel
end trace frame: 0x3, count: 250
0x978bdd844a:
End of stack trace.
 

Re: Help with the NET_LOCK()

[David Hill]

On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> I just enabled the NET_LOCK() again and I'm looking for test reports.
> Please go build a kernel from sources or wait for the next snapshot,
> run it and report back.
> 
> If you're looking for some small coding tasks related to the NET_LOCK()
> just do:
> 
>   # sysctl kern.splassert=2
>   # sysctl kern.pool_debug=2
>   
> Then watch for the traces on your console.
> 
> You'll see something like:
> 
>   Starting stack trace...
>   yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
>   yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
>   pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
>   m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
>   doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
>   sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
>   syscall() at syscall+0x250
>   
> This means accept(2) is doing a memory allocation that can sleep, here
> with m_get(9), while holding the NET_LOCK().  Even if these should be
> ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> be allocated beforehand or simply use the stack for that.
> 
> Cheers,
> Martin
>

splassert: yield: want 0 have 1
Starting stack trace...
yield() at yield+0xac
malloc() at malloc+0x406
ip6_setmoptions() at ip6_setmoptions+0x65
ip6_ctloutput() at ip6_ctloutput+0x6d9
sosetopt() at sosetopt+0x8e
sys_setsockopt() at sys_setsockopt+0x12d
syscall() at syscall+0x27b
--- syscall (number 105) ---
end of kernel
end trace frame: 0x1b1987be0028, count: 250
0x1b197d24b80a:
End of stack trace.
 

Re: Help with the NET_LOCK() - socreate

[David Hill]

On Wed, Jan 25, 2017 at 04:32:25PM +1000, Martin Pieuchot wrote:
> I just enabled the NET_LOCK() again and I'm looking for test reports.
> Please go build a kernel from sources or wait for the next snapshot,
> run it and report back.
> 
> If you're looking for some small coding tasks related to the NET_LOCK()
> just do:
> 
>   # sysctl kern.splassert=2
>   # sysctl kern.pool_debug=2
>   
> Then watch for the traces on your console.
> 
> You'll see something like:
> 
>   Starting stack trace...
>   yield(0,1,d09dac52,f5549dbc,d94e9378) at yield+0xa4
>   yield(d0bc8f40,1,f5549e18,80,14) at yield+0xa4
>   pool_get(d0bc8f40,1,f5549ec8,d03ecbfb,d97815f4) at pool_get+0x1ba
>   m_get(1,3,f5549ec0,d03a9362,d0bc22e0) at m_get+0x30
>   doaccept(d977e6c4,3,cf7ee4f8,cf7ee4ec,2000) at doaccept+0x193
>   sys_accept(d977e6c4,f5549f5c,f5549f7c,0,f5549fa8) at sys_accept+0x37
>   syscall() at syscall+0x250
>   
> This means accept(2) is doing a memory allocation that can sleep, here
> with m_get(9), while holding the NET_LOCK().  Even if these should be
> ok, it is easy to avoid them.  In the case of doaccept() a mbuf could
> be allocated beforehand or simply use the stack for that.
> 
> Cheers,
> Martin
>

splassert: yield: want 0 have 1
Starting stack trace...
yield() at yield+0xac
pool_get() at pool_get+0x1ca
socreate() at socreate+0xba
sys_socket() at sys_socket+0x135
syscall() at syscall+0x27b
--- syscall (number 97) ---
end of kernel
end trace frame: 0x1b1a16c05800, count: 252
0x1b197d23277a:
End of stack trace.

Re: bluhm's splsoftassert

[David Hill]

On Tue, Sep 20, 2016 at 04:17:37PM +0200, Mike Belopuhov wrote:
> On 20 September 2016 at 15:55, Alexander Bluhm <alexander.bl...@gmx.net> 
> wrote:
> > On Tue, Sep 20, 2016 at 08:21:55AM -0400, David Hill wrote:
> >> With bluhm's r1.160 uipc_socket.c.
> >
> > With splsoftnet() in soshutdown() I can fix this one.
> >
> > splassert: sowwakeup: want 5 have 0
> > Starting stack trace...
> > splassert_check() at splassert_check+0x78
> > sowwakeup() at sowwakeup+0x27
> > uipc_usrreq() at uipc_usrreq+0xfd
> > sys_shutdown() at sys_shutdown+0x67
> > syscall() at syscall+0x27b
> > --- syscall (number 134) ---
> > end of kernel
> > end trace frame: 0xe8f2cba5e80, count: 252
> > 0xe8f305dc16a:
> > End of stack trace.
> >
> > ok?
> >
> 
> OK mikeb
> 
> Can we assert that *_usrreq is always called under splsoftnet?
> I recall fixing some of them for raw sockets and some such and
> was wondering if the spl has to be raised before we end up there.
>

Just playing around with splsoftassert's and had to add the following
splsoftnet's to get my machine to be quiet. 

Index: kern/sys_socket.c
===
RCS file: /cvs/src/sys/kern/sys_socket.c,v
retrieving revision 1.22
diff -u -p -r1.22 sys_socket.c
--- kern/sys_socket.c   6 Oct 2016 17:02:10 -   1.22
+++ kern/sys_socket.c   18 Oct 2016 20:15:47 -
@@ -167,6 +167,7 @@ int
 soo_stat(struct file *fp, struct stat *ub, struct proc *p)
 {
struct socket *so = fp->f_data;
+   int s;
 
memset(ub, 0, sizeof (*ub));
ub->st_mode = S_IFSOCK;
@@ -177,8 +178,10 @@ soo_stat(struct file *fp, struct stat *u
ub->st_mode |= S_IWUSR | S_IWGRP | S_IWOTH;
ub->st_uid = so->so_euid;
ub->st_gid = so->so_egid;
+   s = splsoftnet();
(void) ((*so->so_proto->pr_usrreq)(so, PRU_SENSE,
(struct mbuf *)ub, NULL, NULL, p));
+   splx(s);
return (0);
 }
 
Index: kern/uipc_syscalls.c
===
RCS file: /cvs/src/sys/kern/uipc_syscalls.c,v
retrieving revision 1.135
diff -u -p -r1.135 uipc_syscalls.c
--- kern/uipc_syscalls.c8 Oct 2016 02:16:43 -   1.135
+++ kern/uipc_syscalls.c18 Oct 2016 20:15:47 -
@@ -1049,7 +1049,7 @@ sys_getsockname(struct proc *p, void *v,
struct socket *so;
struct mbuf *m = NULL;
socklen_t len;
-   int error;
+   int error, s;
 
if ((error = getsock(p, SCARG(uap, fdes), )) != 0)
return (error);
@@ -1061,7 +1061,9 @@ sys_getsockname(struct proc *p, void *v,
if (error)
goto bad;
m = m_getclr(M_WAIT, MT_SONAME);
+   s = splsoftnet();
error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, 0, m, 0, p);
+   splx(s);
if (error)
goto bad;
error = copyaddrout(p, m, SCARG(uap, asa), len, SCARG(uap, alen));
@@ -1087,7 +1089,7 @@ sys_getpeername(struct proc *p, void *v,
struct socket *so;
struct mbuf *m = NULL;
socklen_t len;
-   int error;
+   int error, s;
 
if ((error = getsock(p, SCARG(uap, fdes), )) != 0)
return (error);
@@ -1103,7 +1105,9 @@ sys_getpeername(struct proc *p, void *v,
if (error)
goto bad;
m = m_getclr(M_WAIT, MT_SONAME);
+   s = splsoftnet();
error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, 0, m, 0, p);
+   splx(s);
if (error)
goto bad;
error = copyaddrout(p, m, SCARG(uap, asa), len, SCARG(uap, alen));
Index: net/if.c
===
RCS file: /cvs/src/sys/net/if.c,v
retrieving revision 1.455
diff -u -p -r1.455 if.c
--- net/if.c16 Oct 2016 21:45:17 -  1.455
+++ net/if.c18 Oct 2016 20:15:47 -
@@ -2072,9 +2072,11 @@ ifioctl(struct socket *so, u_long cmd, c
default:
if (so->so_proto == 0)
return (EOPNOTSUPP);
+   s = splsoftnet();
error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
(struct mbuf *) cmd, (struct mbuf *) data,
(struct mbuf *) ifp, p));
+   splx(s);
break;
}
 
Index: net/pfkey.c
===
RCS file: /cvs/src/sys/net/pfkey.c,v
retrieving revision 1.31
diff -u -p -r1.31 pfkey.c
--- net/pfkey.c 5 Sep 2016 15:12:30 -   1.31
+++ net/pfkey.c 18 Oct 2016 20:15:47 -
@@ -242,6 +242,8 @@ pfkey_usrreq(struct socket *socket, int 
 {
int rval;
 
+   splsoftassert(IPL_SOFTNET);
+
if ((socket->so_proto->pr_protocol > PFKEY_PROTOCOL_MAX) ||

patch: netstat -P - mention allowkmem

[David Hill]

Hello -

netstat -P now requires kern.allowkmem to be set.

Index: netstat.1
===
RCS file: /cvs/src/usr.bin/netstat/netstat.1,v
retrieving revision 1.79
diff -u -p -r1.79 netstat.1
--- netstat.1   1 Sep 2016 14:20:13 -   1.79
+++ netstat.1   12 Oct 2016 03:37:53 -
@@ -234,6 +234,14 @@ option, also print socket, domain and pr
 Only the super-user can use the
 .Fl P
 option.
+.Pp
+The
+.Fl P
+option requires the ability to open
+.Pa /dev/kmem
+which may be restricted based upon the value of the
+.Ar kern.allowkmem
+.Xr sysctl 8 .
 .It Fl p Ar protocol
 Restrict the output to
 .Ar protocol ,

Enable TCP ECN by default (for inbound)

[David Hill]

Hello -

Many operating systems out there are starting to enable TCP ECN by
default, most inbound-only.  Linux, FreeBSD, Apple MacOS and iOS.

This diff mimicks the FreeBSD diff found at
http://marc.info/?l=freebsd-commits-all=146369644516347=2

It modifies the net.inet.tcp.ecn from an off and on switch to a 
three-way setting: off, on, and inbound-only (0, 1, 2, respectively).

It also enables TCP ECN by default for inbound connections that request
it.

My home router has TCP ECN enabled. With 36 hours of uptime, netstat -s
shows:
16165 ECN connections accepted
1 ECE packet received
6 ECE packets sent
589 CWR packets received
109 CWR packets sent

This diff is just to see if there is interest in such a change.  If
people seem to like the idea, perhaps I could update the diff to also:
- Remove TCP_ECN kernel option and always compile in support?
- Only allow 0, 1, or 2 via the sysctl net.inet.tcp.ecn knob.
- Update pf.os

Thoughts?
 - David
 
Index: netinet/tcp_output.c
===
RCS file: /cvs/src/sys/netinet/tcp_output.c,v
retrieving revision 1.118
diff -u -p -r1.118 tcp_output.c
--- netinet/tcp_output.c19 Jul 2016 21:28:43 -  1.118
+++ netinet/tcp_output.c8 Oct 2016 02:30:07 -
@@ -845,10 +845,12 @@ send:
}
if (!(tp->t_flags & TF_DISABLE_ECN)) {
/*
-* if this is a SYN seg, set ECE and CWR.
-* set only ECE for SYN-ACK if peer supports ECN.
+* If net.inet.tcp.ecn is set to 1 and this is a
+* SYN seg, set both ECE and CWR.
+* Set only ECE for SYN-ACK if peer supports ECN.
 */
-   if ((flags & (TH_SYN|TH_ACK)) == TH_SYN)
+   if (tcp_do_ecn == 1 &&
+   (flags & (TH_SYN|TH_ACK)) == TH_SYN)
flags |= (TH_ECE|TH_CWR);
else if ((tp->t_flags & TF_ECN_PERMIT) &&
 (flags & (TH_SYN|TH_ACK)) == (TH_SYN|TH_ACK))
Index: netinet/tcp_subr.c
===
RCS file: /cvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.156
diff -u -p -r1.156 tcp_subr.c
--- netinet/tcp_subr.c  24 Sep 2016 14:51:37 -  1.156
+++ netinet/tcp_subr.c  8 Oct 2016 02:30:07 -
@@ -110,7 +110,7 @@ int tcp_do_sack = 1;/* RFC 2018 selecti
 #endif
 inttcp_ack_on_push = 0;/* set to enable immediate ACK-on-PUSH */
 #ifdef TCP_ECN
-inttcp_do_ecn = 0; /* RFC3168 ECN enabled/disabled? */
+inttcp_do_ecn = 2; /* RFC3168 ECN enabled on inbound requests */
 #endif
 inttcp_do_rfc3390 = 2; /* Increase TCP's Initial Window to 10*mss */
 
Index: netinet/tcp_var.h
===
RCS file: /cvs/src/sys/netinet/tcp_var.h,v
retrieving revision 1.116
diff -u -p -r1.116 tcp_var.h
--- netinet/tcp_var.h   4 Oct 2016 13:54:32 -   1.116
+++ netinet/tcp_var.h   8 Oct 2016 02:30:07 -
@@ -579,7 +579,7 @@ extern  int tcp_do_sack;/* SACK enabled/
 extern struct pool sackhl_pool;
 extern int tcp_sackhole_limit; /* max entries for tcp sack queues */
 #endif
-extern int tcp_do_ecn; /* RFC3168 ECN enabled/disabled? */
+extern int tcp_do_ecn; /* RFC3168 ECN: 0:off 1:on 2:inbound-only */
 extern int tcp_do_rfc3390; /* RFC3390 Increasing TCP's Initial Window */
 
 extern struct pool tcpqe_pool;

Re: Help me testing the netlock

[David Hill]

Hi Martin -

As soon as machine boots and first packet hits the VPN, it hits the
assertion:
rw_status() == RW_WRITE

(hand-typed)

TID PID UID PRFLAGS PFLAGS  CPU COMMAND
*78967  78967   0   0x14000 0x200   1   crypto

__assert at0x25
ip_output() at 0x843
ipsp_process_done() at 0x2ad
esp_output_cb at 0x135
taskq_thnread at 0x6c

Re: my final netinet6 bcopy->memcpy

[David Hill]

Ping.

On Mon, Sep 19, 2016 at 09:22:50PM -0400, David Hill wrote:
> Hello -
> 
> Here are the final bcopy->memcpy conversions in netinet6 that I am
> comfortable with.  There is also one (bcmp()) to (memcmp() != 0)
> conversion since the memory is not overlapping.
> 
> This, with the last netinet6 diff, has dropped the number of bcopy()
> calls from 62 to 29 in netinet6/
> 
> Index: icmp6.c
> ===
> RCS file: /cvs/src/sys/netinet6/icmp6.c,v
> retrieving revision 1.190
> diff -u -p -r1.190 icmp6.c
> --- icmp6.c   24 Aug 2016 09:38:29 -  1.190
> +++ icmp6.c   20 Sep 2016 01:00:36 -
> @@ -1401,7 +1401,7 @@ icmp6_redirect_input(struct mbuf *m, int
>   bzero(, sizeof(sin6));
>   sin6.sin6_family = AF_INET6;
>   sin6.sin6_len = sizeof(struct sockaddr_in6);
> - bcopy(, _addr, sizeof(reddst6));
> + memcpy(_addr, , sizeof(reddst6));
>   rt = rtalloc(sin6tosa(), 0, m->m_pkthdr.ph_rtableid);
>   if (rt) {
>   if (rt->rt_gateway == NULL ||
> @@ -1509,9 +1509,9 @@ icmp6_redirect_input(struct mbuf *m, int
>   sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = 
> AF_INET6;
>   sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
>   sizeof(struct sockaddr_in6);
> - bcopy(, _addr, sizeof(struct in6_addr));
> - bcopy(, _addr, sizeof(struct in6_addr));
> - bcopy(, _addr, sizeof(struct in6_addr));
> + memcpy(_addr, , sizeof(struct in6_addr));
> + memcpy(_addr, , sizeof(struct in6_addr));
> + memcpy(_addr, , sizeof(struct in6_addr));
>   rtredirect(sin6tosa(), sin6tosa(), sin6tosa(),
>   , m->m_pkthdr.ph_rtableid);
>  
> @@ -1528,7 +1528,7 @@ icmp6_redirect_input(struct mbuf *m, int
>   bzero(, sizeof(sdst));
>   sdst.sin6_family = AF_INET6;
>   sdst.sin6_len = sizeof(struct sockaddr_in6);
> - bcopy(, _addr, sizeof(struct in6_addr));
> + memcpy(_addr, , sizeof(struct in6_addr));
>   pfctlinput(PRC_REDIRECT_HOST, sin6tosa());
>   }
>  
> Index: in6.c
> ===
> RCS file: /cvs/src/sys/netinet6/in6.c,v
> retrieving revision 1.192
> diff -u -p -r1.192 in6.c
> --- in6.c 4 Sep 2016 10:32:01 -   1.192
> +++ in6.c 20 Sep 2016 01:00:36 -
> @@ -1029,9 +1029,9 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
>  
>   /* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
>   bzero(, sizeof(ifra));
> - bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
> + memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
>  
> - bcopy(>addr, _addr,
> + memcpy(_addr, >addr,
>   ((struct sockaddr *)>addr)->sa_len);
>   if (hostid) {
>   /* fill in hostid part */
> @@ -1042,7 +1042,7 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
>   }
>  
>   if (((struct sockaddr *)>dstaddr)->sa_family) {   /*XXX*/
> - bcopy(>dstaddr, _dstaddr,
> + memcpy(_dstaddr, >dstaddr,
>   ((struct sockaddr *)>dstaddr)->sa_len);
>   if (hostid) {
>   ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
> @@ -1073,14 +1073,14 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
>   in6_prefixlen2mask(, iflr->prefixlen);
>  
>   sin6 = (struct sockaddr_in6 *)>addr;
> - bcopy(>sin6_addr, , sizeof(match));
> + memcpy(, >sin6_addr, sizeof(match));
>   match.s6_addr32[0] &= mask.s6_addr32[0];
>   match.s6_addr32[1] &= mask.s6_addr32[1];
>   match.s6_addr32[2] &= mask.s6_addr32[2];
>   match.s6_addr32[3] &= mask.s6_addr32[3];
>  
>   /* if you set extra bits, that's wrong */
> - if (bcmp(, >sin6_addr, sizeof(match)))
> + if (memcmp(, >sin6_addr, sizeof(match)) != 
> 0)
>   return EINVAL;
>  
>   cmp = 1;
> @@ -1092,7 +1092,7 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
>   /* on deleting an address, do exact match */
>   in6_prefixlen2mask(, 128);
>   sin6 = (struct sockaddr_in6 *)>addr;
> -  

Re: splassert: sorwakeup: want 5 have 4

[David Hill]

On Mon, Sep 26, 2016 at 01:55:01PM +0100, Stuart Henderson wrote:
> Seen about a dozen times on a box doing ospfd + ospf6d (for internal
> routes), bgpd (for default route) + isakmpd, shortly after startup
> after updating:
> 
> splassert: sorwakeup: want 5 have 4   
> Starting stack trace...  
> splassert_check() at splassert_check+0x78
> sorwakeup() at sorwakeup+0x27
> timeout_run() at timeout_run+0x48
> softclock() at softclock+0x14c   
> softintr_dispatch() at softintr_dispatch+0x8b
> Xsoftclock() at Xsoftclock+0x1f  
> --- interrupt ---  
> end of kernel
> end trace frame: 0x51, count: 251
> 0x8: 
> End of stack trace.
>

Perhaps this?

Index: rtsock.c
===
RCS file: /cvs/src/sys/net/rtsock.c,v
retrieving revision 1.206
diff -u -p -r1.206 rtsock.c
--- rtsock.c24 Sep 2016 19:27:10 -  1.206
+++ rtsock.c26 Sep 2016 17:23:21 -
@@ -302,6 +302,7 @@ rt_senddesync(void *data)
struct rawcb*rp;
struct routecb  *rop;
struct mbuf *desync_mbuf;
+   int  s;
 
rp = (struct rawcb *)data;
rop = (struct routecb *)rp;
@@ -317,7 +318,9 @@ rt_senddesync(void *data)
(sbappendaddr(>rcb_socket->so_rcv, _src, 
desync_mbuf, (struct mbuf *)NULL) != 0)) {
rop->flags &= ~ROUTECB_FLAG_DESYNC;
+   s = splsoftnet();
sorwakeup(rp->rcb_socket);
+   splx(s);
} else {
m_freem(desync_mbuf);
/* Re-add timeout to try sending msg again */

Re: bluhm's splsoftassert

[David Hill]

On Wed, Sep 21, 2016 at 12:32:50PM -0400, David Hill wrote:
> Hit another.
> 
> splassert: sorwakeup: want 5 have 0
> Starting stack trace...
> splassert_check() at splassert_check+0x78
> sorwakeup() at sorwakeup+0x27
> route_input() at route_input+0x284
> ifioctl() at ifioctl+0x765
> sys_ioctl() at sys_ioctl+0x196
> syscall() at syscall+0x27b
> --- syscall (number 54) ---
> end of kernel
> end trace frame: 0x7f7e72f0, count: 251
> 0x10044651af1a:
> End of stack trace.
>

Index: if.c
===
RCS file: /cvs/src/sys/net/if.c,v
retrieving revision 1.449
diff -u -p -r1.449 if.c
--- if.c20 Sep 2016 16:14:43 -  1.449
+++ if.c21 Sep 2016 16:51:39 -
@@ -1898,7 +1898,9 @@ ifioctl(struct socket *so, u_long cmd, c
 
ifp->if_xflags = (ifp->if_xflags & IFXF_CANTCHANGE) |
(ifr->ifr_flags & ~IFXF_CANTCHANGE);
+   s = splsoftnet();
rt_ifmsg(ifp);
+   splx(s);
break;
 
case SIOCSIFMETRIC:

Re: bluhm's splsoftassert

[David Hill]

Hit another.

splassert: sorwakeup: want 5 have 0
Starting stack trace...
splassert_check() at splassert_check+0x78
sorwakeup() at sorwakeup+0x27
route_input() at route_input+0x284
ifioctl() at ifioctl+0x765
sys_ioctl() at sys_ioctl+0x196
syscall() at syscall+0x27b
--- syscall (number 54) ---
end of kernel
end trace frame: 0x7f7e72f0, count: 251
0x10044651af1a:
End of stack trace.

remove splsoftnet from *_clone_destroy

[David Hill]

Hello -

I believe this is correct.

After mikeb@ committed destroying cloneable interfaces under
splsoftnet, these aren't needed anymore.

Index: if_mpe.c
===
RCS file: /cvs/src/sys/net/if_mpe.c,v
retrieving revision 1.54
diff -u -p -r1.54 if_mpe.c
--- if_mpe.c13 Apr 2016 11:41:15 -  1.54
+++ if_mpe.c21 Sep 2016 01:30:27 -
@@ -117,15 +117,12 @@ int
 mpe_clone_destroy(struct ifnet *ifp)
 {
struct mpe_softc*mpeif = ifp->if_softc;
-   int s;
 
LIST_REMOVE(mpeif, sc_list);
 
if (mpeif->sc_smpls.smpls_label) {
-   s = splsoftnet();
rt_ifa_del(>sc_ifa, RTF_MPLS,
smplstosa(>sc_smpls));
-   splx(s);
}
 
if_detach(ifp);
Index: if_mpw.c
===
RCS file: /cvs/src/sys/net/if_mpw.c,v
retrieving revision 1.14
diff -u -p -r1.14 if_mpw.c
--- if_mpw.c13 Apr 2016 11:41:15 -  1.14
+++ if_mpw.c21 Sep 2016 01:30:27 -
@@ -120,15 +120,12 @@ int
 mpw_clone_destroy(struct ifnet *ifp)
 {
struct mpw_softc *sc = ifp->if_softc;
-   int s;
 
ifp->if_flags &= ~IFF_RUNNING;
 
if (sc->sc_smpls.smpls_label) {
-   s = splsoftnet();
rt_ifa_del(>sc_ifa, RTF_MPLS,
smplstosa(>sc_smpls));
-   splx(s);
}
 
if_ih_remove(ifp, mpw_input, NULL);
Index: if_pfsync.c
===
RCS file: /cvs/src/sys/net/if_pfsync.c,v
retrieving revision 1.231
diff -u -p -r1.231 if_pfsync.c
--- if_pfsync.c 15 Sep 2016 02:00:18 -  1.231
+++ if_pfsync.c 21 Sep 2016 01:30:27 -
@@ -353,9 +353,7 @@ pfsync_clone_destroy(struct ifnet *ifp)
 {
struct pfsync_softc *sc = ifp->if_softc;
struct pfsync_deferral *pd;
-   int s;
 
-   s = splsoftnet();
timeout_del(>sc_bulkfail_tmo);
timeout_del(>sc_bulk_tmo);
timeout_del(>sc_tmo);
@@ -384,7 +382,6 @@ pfsync_clone_destroy(struct ifnet *ifp)
free(sc, M_DEVBUF, sizeof(*sc));
 
pfsyncif = NULL;
-   splx(s);
 
return (0);
 }

Re: bluhm's splsoftassert

[David Hill]

On Tue, Sep 20, 2016 at 05:04:02PM +0200, Mike Belopuhov wrote:
> On Tue, Sep 20, 2016 at 10:51 -0400, David Hill wrote:
> > On Tue, Sep 20, 2016 at 09:53:02AM -0400, David Hill wrote:
> > > More...
> > > 
> > > splassert: sorwakeup: want 5 have 0
> > > Starting stack trace...
> > > splassert_check() at splassert_check+0x78
> > > sorwakeup() at sorwakeup+0x27
> > > route_input() at route_input+0x284
> > > pflog_clone_create() at pflog_clone_create+0xa4
> > > if_clone_create() at if_clone_create+0x7f
> > > ifioctl() at ifioctl+0x35a
> > > sys_ioctl() at sys_ioctl+0x196
> > > syscall() at syscall+0x27b
> > > --- syscall (number 54) ---
> > > end of kernel
> > > end trace frame: 0x7f7c7930, count: 249
> > > 0x1aeaedf1af1a:
> > > End of stack trace.
> > > 
> > >
> > 
> > Another similar...
> > 
> > Sep 20 10:33:25 olive /bsd: splassert: sorwakeup: want 5 have 0
> > Sep 20 10:33:25 olive /bsd: Starting stack trace...
> > Sep 20 10:33:25 olive /bsd: splassert_check() at splassert_check+0x78
> > Sep 20 10:33:25 olive /bsd: sorwakeup() at sorwakeup+0x27
> > Sep 20 10:33:25 olive /bsd: route_input() at route_input+0x284
> > Sep 20 10:33:25 olive /bsd: vether_clone_create() at
> > vether_clone_create+0xd9
> > Sep 20 10:33:25 olive /bsd: if_clone_create() at if_clone_create+0x7f
> > Sep 20 10:33:25 olive /bsd: ifioctl() at ifioctl+0x35a
> > Sep 20 10:33:25 olive /bsd: sys_ioctl() at sys_ioctl+0x196
> > Sep 20 10:33:25 olive /bsd: syscall() at syscall+0x27b
> > Sep 20 10:33:25 olive /bsd: --- syscall (number 54) ---
> > Sep 20 10:33:25 olive /bsd: end of kernel
> > Sep 20 10:33:25 olive /bsd: end trace frame: 0x7f7d0df1, count: 249
> > Sep 20 10:33:25 olive /bsd: 0x119ca1c1af1a:
> > Sep 20 10:33:25 olive /bsd: End of stack trace.
> >  
> > 
> 
> It's all the same.  I'm not certain what would be the best way to
> go around this, but a cautious approach would be something like this:
> just wrapping ifc_create, ifc_destroy in splsoftnet.
> 
> David, can you give this a spin?
> 
> Index: sys/net/if.c
> ===
> RCS file: /home/cvs/src/sys/net/if.c,v
> retrieving revision 1.448
> diff -u -p -r1.448 if.c
> --- sys/net/if.c  13 Sep 2016 08:15:01 -  1.448
> +++ sys/net/if.c  20 Sep 2016 14:58:57 -
> @@ -1041,7 +1041,7 @@ if_clone_create(const char *name, int rd
>  {
>   struct if_clone *ifc;
>   struct ifnet *ifp;
> - int unit, ret;
> + int unit, ret, s;
>  
>   ifc = if_clone_lookup(name, );
>   if (ifc == NULL)
> @@ -1050,9 +1050,13 @@ if_clone_create(const char *name, int rd
>   if (ifunit(name) != NULL)
>   return (EEXIST);
>  
> + s = splsoftnet();
>   if ((ret = (*ifc->ifc_create)(ifc, unit)) != 0 ||
> - (ifp = ifunit(name)) == NULL)
> + (ifp = ifunit(name)) == NULL) {
> + splx(s);
>   return (ret);
> + }
> + splx(s);
>  
>   if_addgroup(ifp, ifc->ifc_name);
>   if (rdomain != 0)
> @@ -1069,7 +1073,7 @@ if_clone_destroy(const char *name)
>  {
>   struct if_clone *ifc;
>   struct ifnet *ifp;
> - int s;
> + int error, s;
>  
>   ifc = if_clone_lookup(name, NULL);
>   if (ifc == NULL)
> @@ -1088,7 +1092,10 @@ if_clone_destroy(const char *name)
>   splx(s);
>   }
>  
> - return ((*ifc->ifc_destroy)(ifp));
> + s = splsoftnet();
> + error = (*ifc->ifc_destroy)(ifp);
> + splx(s);
> + return (error);
>  }
>  
>  /*
>

Works.  No more assert and functions normally. 

Re: bluhm's splsoftassert

[David Hill]

On Tue, Sep 20, 2016 at 09:53:02AM -0400, David Hill wrote:
> More...
> 
> splassert: sorwakeup: want 5 have 0
> Starting stack trace...
> splassert_check() at splassert_check+0x78
> sorwakeup() at sorwakeup+0x27
> route_input() at route_input+0x284
> pflog_clone_create() at pflog_clone_create+0xa4
> if_clone_create() at if_clone_create+0x7f
> ifioctl() at ifioctl+0x35a
> sys_ioctl() at sys_ioctl+0x196
> syscall() at syscall+0x27b
> --- syscall (number 54) ---
> end of kernel
> end trace frame: 0x7f7c7930, count: 249
> 0x1aeaedf1af1a:
> End of stack trace.
> 
>

Another similar...

Sep 20 10:33:25 olive /bsd: splassert: sorwakeup: want 5 have 0
Sep 20 10:33:25 olive /bsd: Starting stack trace...
Sep 20 10:33:25 olive /bsd: splassert_check() at splassert_check+0x78
Sep 20 10:33:25 olive /bsd: sorwakeup() at sorwakeup+0x27
Sep 20 10:33:25 olive /bsd: route_input() at route_input+0x284
Sep 20 10:33:25 olive /bsd: vether_clone_create() at
vether_clone_create+0xd9
Sep 20 10:33:25 olive /bsd: if_clone_create() at if_clone_create+0x7f
Sep 20 10:33:25 olive /bsd: ifioctl() at ifioctl+0x35a
Sep 20 10:33:25 olive /bsd: sys_ioctl() at sys_ioctl+0x196
Sep 20 10:33:25 olive /bsd: syscall() at syscall+0x27b
Sep 20 10:33:25 olive /bsd: --- syscall (number 54) ---
Sep 20 10:33:25 olive /bsd: end of kernel
Sep 20 10:33:25 olive /bsd: end trace frame: 0x7f7d0df1, count: 249
Sep 20 10:33:25 olive /bsd: 0x119ca1c1af1a:
Sep 20 10:33:25 olive /bsd: End of stack trace.
 

Re: bluhm's splsoftassert

[David Hill]

More...

splassert: sorwakeup: want 5 have 0
Starting stack trace...
splassert_check() at splassert_check+0x78
sorwakeup() at sorwakeup+0x27
route_input() at route_input+0x284
pflog_clone_create() at pflog_clone_create+0xa4
if_clone_create() at if_clone_create+0x7f
ifioctl() at ifioctl+0x35a
sys_ioctl() at sys_ioctl+0x196
syscall() at syscall+0x27b
--- syscall (number 54) ---
end of kernel
end trace frame: 0x7f7c7930, count: 249
0x1aeaedf1af1a:
End of stack trace.

Re: bluhm's splsoftassert

[David Hill]

Another

splassert: sowwakeup: want 5 have 0
Starting stack trace...
splassert_check() at splassert_check+0x78
sowwakeup() at sowwakeup+0x27
uipc_usrreq() at uipc_usrreq+0xfd
sys_shutdown() at sys_shutdown+0x67
syscall() at syscall+0x27b
--- syscall (number 134) ---
end of kernel
end trace frame: 0xfd39b922600, count: 252
0xfd2c6bbcffa:
End of stack trace.

Re: bluhm's splsoftassert

[David Hill]

On Tue, Sep 20, 2016 at 03:16:50PM +0200, Alexander Bluhm wrote:
> On Tue, Sep 20, 2016 at 08:21:55AM -0400, David Hill wrote:
> > With bluhm's r1.160 uipc_socket.c.
> > Here are the first ones that were detected.
> 
> Thanks for the fast report.
> 
> So fifo works around the socket layer.  Better call soconnect2()
> instead of unp_connect2().  This adds the missing splsoftnet().
> 
> I think we should demand that socantsendmore() and socantrcvmore()
> in uipc_socket2.c should be called with splsoftnet().

Should both socantsendmore() and socantrcvmore() get a splsoftassert()
then, for now? 

> 
> ok?
> 
> bluhm
> 
> Index: miscfs/fifofs/fifo_vnops.c
> ===
> RCS file: /data/mirror/openbsd/cvs/src/sys/miscfs/fifofs/fifo_vnops.c,v
> retrieving revision 1.51
> diff -u -p -r1.51 fifo_vnops.c
> --- miscfs/fifofs/fifo_vnops.c7 Jun 2016 06:12:37 -   1.51
> +++ miscfs/fifofs/fifo_vnops.c20 Sep 2016 13:04:56 -
> @@ -48,7 +48,6 @@
>  #include 
>  #include 
>  #include 
> -#include 
>  #include 
>  
>  #include 
> @@ -143,7 +142,7 @@ fifo_open(void *v)
>   return (error);
>   }
>   fip->fi_writesock = wso;
> - if ((error = unp_connect2(wso, rso)) != 0) {
> + if ((error = soconnect2(wso, rso)) != 0) {
>   (void)soclose(wso);
>   (void)soclose(rso);
>   free(fip, M_VNODE, sizeof *fip);
> @@ -350,20 +349,25 @@ fifo_close(void *v)
>   struct vop_close_args *ap = v;
>   struct vnode *vp = ap->a_vp;
>   struct fifoinfo *fip = vp->v_fifoinfo;
> - int error1 = 0, error2 = 0;
> + int s, error1 = 0, error2 = 0;
>  
>   if (fip == NULL)
>   return (0);
>  
>   if (ap->a_fflag & FREAD) {
> - if (--fip->fi_readers == 0)
> + if (--fip->fi_readers == 0) {
> + s = splsoftnet();
>   socantsendmore(fip->fi_writesock);
> + splx(s);
> + }
>   }
>   if (ap->a_fflag & FWRITE) {
>   if (--fip->fi_writers == 0) {
> + s = splsoftnet();
>   /* SS_ISDISCONNECTED will result in POLLHUP */
>   fip->fi_readsock->so_state |= SS_ISDISCONNECTED;
>   socantrcvmore(fip->fi_readsock);
> + splx(s);
>   }
>   }
>   if (fip->fi_readers == 0 && fip->fi_writers == 0) {
> 

Re: bluhm's splsoftassert

[David Hill]

And another.

splassert: sorwakeup: want 5 have 4
Starting stack trace...
splassert_check() at splassert_check+0x78
sorwakeup() at sorwakeup+0x27
pfkey_sendup() at pfkey_sendup+0x99
pfkeyv2_sendmessage() at pfkeyv2_sendmessage+0x226
pfkeyv2_expire() at pfkeyv2_expire+0x18d
tdb_soft_timeout() at tdb_soft_timeout+0x1c
softclock() at softclock+0x144
softintr_dispatch() at softintr_dispatch+0x8b
Xsoftclock() at Xsoftclock+0x1f
--- interrupt ---
end of kernel
end trace frame: 0x20, count: 248
0x8:
End of stack trace.

Re: bluhm's splsoftassert

[David Hill]

Another...

splassert: sorwakeup: want 5 have 4
Starting stack trace...
splassert_check() at splassert_check+0x78
sorwakeup() at sorwakeup+0x27
pfkey_sendup() at pfkey_sendup+0x99
pfkeyv2_sendmessage() at pfkeyv2_sendmessage+0x226
pfkeyv2_expire() at pfkeyv2_expire+0x18d
tdb_timeout() at tdb_timeout+0x2f
softclock() at softclock+0x144
softintr_dispatch() at softintr_dispatch+0x8b
Xsoftclock() at Xsoftclock+0x1f
--- interrupt ---
end of kernel
end trace frame: 0x20, count: 248
0x8:
End of stack trace.

bluhm's splsoftassert

[David Hill]

Hello -

With bluhm's r1.160 uipc_socket.c.

Here are the first ones that were detected.

splassert: sowwakeup: want 5 have 0
Starting stack trace...
splassert_check() at splassert_check+0x78
sowwakeup() at sowwakeup+0x27
unp_connect2() at unp_connect2+0x62
fifo_open() at fifo_open+0x244
VOP_OPEN() at VOP_OPEN+0x3f
vn_open() at vn_open+0x16f
doopenat() at doopenat+0x187
syscall() at syscall+0x27b
--- syscall (number 5) ---
end of kernel
end trace frame: 0x199f2ba48898, count: 249
0x199ebb5c6d0a:
End of stack trace.
splassert: sorwakeup: want 5 have 0
Starting stack trace...
splassert_check() at splassert_check+0x78
sorwakeup() at sorwakeup+0x27
soisconnected() at soisconnected+0x46
unp_connect2() at unp_connect2+0x5a
fifo_open() at fifo_open+0x244
VOP_OPEN() at VOP_OPEN+0x3f
vn_open() at vn_open+0x16f
doopenat() at doopenat+0x187
syscall() at syscall+0x27b
--- syscall (number 5) ---
end of kernel
end trace frame: 0x199e63a704a8, count: 248
0x199ebb5c6d0a:
End of stack trace.
--- Begin Message ---
CVSROOT:/cvs
Module name:src
Changes by: bl...@cvs.openbsd.org   2016/09/20 05:11:44

Modified files:
sys/kern   : uipc_socket.c 

Log message:
Add some spl softnet assertions that will help us to find the right
places for the upcoming network lock.  This might trigger some
asserts, but we have to find the missing code paths.
OK mpi@

--- End Message ---

netinet/ ansify

[David Hill]

Hello -

Ansify netinet/

Index: ip_ecn.c
===
RCS file: /cvs/src/sys/netinet/ip_ecn.c,v
retrieving revision 1.7
diff -u -p -r1.7 ip_ecn.c
--- ip_ecn.c19 Dec 2014 17:14:40 -  1.7
+++ ip_ecn.c20 Sep 2016 01:54:47 -
@@ -84,10 +84,7 @@
  * call it after you've done the default initialization/copy for the outer.
  */
 void
-ip_ecn_ingress(mode, outer, inner)
-   int mode;
-   u_int8_t *outer;
-   u_int8_t *inner;
+ip_ecn_ingress(int mode, u_int8_t *outer, u_int8_t *inner)
 {
if (!outer || !inner)
panic("NULL pointer passed to ip_ecn_ingress");
@@ -120,10 +117,7 @@ ip_ecn_ingress(mode, outer, inner)
  * the caller should drop the packet if the return value is 0.
  */
 int
-ip_ecn_egress(mode, outer, inner)
-   int mode;
-   u_int8_t *outer;
-   u_int8_t *inner;
+ip_ecn_egress(int mode, u_int8_t *outer, u_int8_t *inner)
 {
if (!outer || !inner)
panic("NULL pointer passed to ip_ecn_egress");
Index: ip_ether.c
===
RCS file: /cvs/src/sys/netinet/ip_ether.c,v
retrieving revision 1.80
diff -u -p -r1.80 ip_ether.c
--- ip_ether.c  2 Dec 2015 08:47:00 -   1.80
+++ ip_ether.c  20 Sep 2016 01:54:47 -
@@ -507,11 +507,8 @@ etherip_output(struct mbuf *m, struct td
 }
 
 int
-etherip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
-   int *name;
-   u_int namelen;
-   void *oldp, *newp;
-   size_t *oldlenp, newlen;
+etherip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
+void *newp, size_t newlen)
 {
/* All sysctl names at this level are terminal. */
if (namelen != 1)
Index: ip_ipcomp.c
===
RCS file: /cvs/src/sys/netinet/ip_ipcomp.c,v
retrieving revision 1.47
diff -u -p -r1.47 ip_ipcomp.c
--- ip_ipcomp.c 13 Sep 2016 19:56:55 -  1.47
+++ ip_ipcomp.c 20 Sep 2016 01:54:47 -
@@ -80,10 +80,7 @@ ipcomp_attach(void)
  * ipcomp_init() is called when an CPI is being set up.
  */
 int
-ipcomp_init(tdbp, xsp, ii)
-   struct tdb *tdbp;
-   struct xformsw *xsp;
-   struct ipsecinit *ii;
+ipcomp_init(struct tdb *tdbp, struct xformsw *xsp, struct ipsecinit *ii)
 {
struct comp_algo *tcomp = NULL;
struct cryptoini cric;
@@ -120,8 +117,7 @@ ipcomp_init(tdbp, xsp, ii)
  * ipcomp_zeroize() used when IPCA is deleted
  */
 int
-ipcomp_zeroize(tdbp)
-   struct tdb *tdbp;
+ipcomp_zeroize(struct tdb *tdbp)
 {
int err;
 
@@ -134,11 +130,7 @@ ipcomp_zeroize(tdbp)
  * ipcomp_input() gets called to uncompress an input packet
  */
 int
-ipcomp_input(m, tdb, skip, protoff)
-   struct mbuf*m;
-   struct tdb *tdb;
-   int skip;
-   int protoff;
+ipcomp_input(struct mbuf *m, struct tdb *tdb, int skip, int protoff)
 {
struct comp_algo *ipcompx = (struct comp_algo *) tdb->tdb_compalgxform;
struct tdb_crypto *tc;
@@ -361,12 +353,8 @@ baddone:
  * IPComp output routine, called by ipsp_process_packet()
  */
 int
-ipcomp_output(m, tdb, mp, skip, protoff)
-   struct mbuf*m;
-   struct tdb *tdb;
-   struct mbuf   **mp;
-   int skip;
-   int protoff;
+ipcomp_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int skip,
+int protoff)
 {
struct comp_algo *ipcompx = (struct comp_algo *) tdb->tdb_compalgxform;
int hlen;
Index: tcp_subr.c
===
RCS file: /cvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.155
diff -u -p -r1.155 tcp_subr.c
--- tcp_subr.c  15 Sep 2016 02:00:18 -  1.155
+++ tcp_subr.c  20 Sep 2016 01:54:47 -
@@ -186,8 +186,7 @@ tcp_init(void)
  * into just an IP overlay pointer, with casting as appropriate for v6. rja
  */
 struct mbuf *
-tcp_template(tp)
-   struct tcpcb *tp;
+tcp_template(struct tcpcb *tp)
 {
struct inpcb *inp = tp->t_inpcb;
struct mbuf *m;
@@ -488,9 +487,7 @@ tcp_newtcpcb(struct inpcb *inp)
  * then send a RST to peer.
  */
 struct tcpcb *
-tcp_drop(tp, errno)
-   struct tcpcb *tp;
-   int errno;
+tcp_drop(struct tcpcb *tp, int errno)
 {
struct socket *so = tp->t_inpcb->inp_socket;
 
@@ -595,9 +592,7 @@ tcp_rscale(struct tcpcb *tp, u_long hiwa
  * (for now, won't do anything until can select for soft error).
  */
 void
-tcp_notify(inp, error)
-   struct inpcb *inp;
-   int error;
+tcp_notify(struct inpcb *inp, int error)
 {
struct tcpcb *tp = intotcpcb(inp);
struct socket *so = inp->inp_socket;
@@ -863,9 +858,7 @@ tcp_ctlinput(int cmd, struct sockaddr *s
  * Path MTU Discovery handlers.
  */
 void
-tcp6_mtudisc_callback(sin6, rdomain)
-   struct sockaddr_in6 *sin6;
-   u_int rdomain;
+tcp6_mtudisc_callback(struct sockaddr_in6 *sin6, u_int rdomain)

my final netinet6 bcopy->memcpy

[David Hill]

Hello -

Here are the final bcopy->memcpy conversions in netinet6 that I am
comfortable with.  There is also one (bcmp()) to (memcmp() != 0)
conversion since the memory is not overlapping.

This, with the last netinet6 diff, has dropped the number of bcopy()
calls from 62 to 29 in netinet6/

Index: icmp6.c
===
RCS file: /cvs/src/sys/netinet6/icmp6.c,v
retrieving revision 1.190
diff -u -p -r1.190 icmp6.c
--- icmp6.c 24 Aug 2016 09:38:29 -  1.190
+++ icmp6.c 20 Sep 2016 01:00:36 -
@@ -1401,7 +1401,7 @@ icmp6_redirect_input(struct mbuf *m, int
bzero(, sizeof(sin6));
sin6.sin6_family = AF_INET6;
sin6.sin6_len = sizeof(struct sockaddr_in6);
-   bcopy(, _addr, sizeof(reddst6));
+   memcpy(_addr, , sizeof(reddst6));
rt = rtalloc(sin6tosa(), 0, m->m_pkthdr.ph_rtableid);
if (rt) {
if (rt->rt_gateway == NULL ||
@@ -1509,9 +1509,9 @@ icmp6_redirect_input(struct mbuf *m, int
sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = 
AF_INET6;
sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
sizeof(struct sockaddr_in6);
-   bcopy(, _addr, sizeof(struct in6_addr));
-   bcopy(, _addr, sizeof(struct in6_addr));
-   bcopy(, _addr, sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
rtredirect(sin6tosa(), sin6tosa(), sin6tosa(),
, m->m_pkthdr.ph_rtableid);
 
@@ -1528,7 +1528,7 @@ icmp6_redirect_input(struct mbuf *m, int
bzero(, sizeof(sdst));
sdst.sin6_family = AF_INET6;
sdst.sin6_len = sizeof(struct sockaddr_in6);
-   bcopy(, _addr, sizeof(struct in6_addr));
+   memcpy(_addr, , sizeof(struct in6_addr));
pfctlinput(PRC_REDIRECT_HOST, sin6tosa());
}
 
Index: in6.c
===
RCS file: /cvs/src/sys/netinet6/in6.c,v
retrieving revision 1.192
diff -u -p -r1.192 in6.c
--- in6.c   4 Sep 2016 10:32:01 -   1.192
+++ in6.c   20 Sep 2016 01:00:36 -
@@ -1029,9 +1029,9 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
 
/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
bzero(, sizeof(ifra));
-   bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
+   memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
 
-   bcopy(>addr, _addr,
+   memcpy(_addr, >addr,
((struct sockaddr *)>addr)->sa_len);
if (hostid) {
/* fill in hostid part */
@@ -1042,7 +1042,7 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
}
 
if (((struct sockaddr *)>dstaddr)->sa_family) {   /*XXX*/
-   bcopy(>dstaddr, _dstaddr,
+   memcpy(_dstaddr, >dstaddr,
((struct sockaddr *)>dstaddr)->sa_len);
if (hostid) {
ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
@@ -1073,14 +1073,14 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
in6_prefixlen2mask(, iflr->prefixlen);
 
sin6 = (struct sockaddr_in6 *)>addr;
-   bcopy(>sin6_addr, , sizeof(match));
+   memcpy(, >sin6_addr, sizeof(match));
match.s6_addr32[0] &= mask.s6_addr32[0];
match.s6_addr32[1] &= mask.s6_addr32[1];
match.s6_addr32[2] &= mask.s6_addr32[2];
match.s6_addr32[3] &= mask.s6_addr32[3];
 
/* if you set extra bits, that's wrong */
-   if (bcmp(, >sin6_addr, sizeof(match)))
+   if (memcmp(, >sin6_addr, sizeof(match)) != 
0)
return EINVAL;
 
cmp = 1;
@@ -1092,7 +1092,7 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
/* on deleting an address, do exact match */
in6_prefixlen2mask(, 128);
sin6 = (struct sockaddr_in6 *)>addr;
-   bcopy(>sin6_addr, , sizeof(match));
+   memcpy(, >sin6_addr, sizeof(match));
 
cmp = 1;
}
@@ -1104,7 +1104,7 @@ in6_lifaddr_ioctl(u_long cmd, caddr_t da
if (!cmp)
break;
 
-   bcopy(IFA_IN6(ifa), , sizeof(candidate));
+   memcpy(, IFA_IN6(ifa), sizeof(candidate));
candidate.s6_addr32[0] &= 

netinet/tcp_input.c syn_cache_get diff

[David Hill]

Hello -

Make sure we keep TF_NOPUSH set if TCP_NOPUSH was set.

FreeBSD has the same:
https://github.com/freebsd/freebsd/blob/c9af4f2541fd437e0805365fbeec46d69e033310/sys/netinet/tcp_syncache.c#L860
 
Index: netinet/tcp_input.c
===
RCS file: /cvs/src/sys/netinet/tcp_input.c,v
retrieving revision 1.327
diff -u -p -r1.327 tcp_input.c
--- netinet/tcp_input.c 15 Sep 2016 02:00:18 -  1.327
+++ netinet/tcp_input.c 18 Sep 2016 17:07:09 -
@@ -3798,7 +3798,7 @@ syn_cache_get(struct sockaddr *src, stru
(void) m_free(am);
 
tp = intotcpcb(inp);
-   tp->t_flags = sototcpcb(oso)->t_flags & TF_NODELAY;
+   tp->t_flags = sototcpcb(oso)->t_flags & (TF_NOPUSH|TF_NODELAY);
if (sc->sc_request_r_scale != 15) {
tp->requested_s_scale = sc->sc_requested_s_scale;
tp->request_r_scale = sc->sc_request_r_scale;

crypto/ bcopy->memcpy

[David Hill]

Hello -

Here are a few bcopy to memcpy conversions for crypto/

Index: crypto/crypto.c
===
RCS file: /cvs/src/sys/crypto/crypto.c,v
retrieving revision 1.77
diff -u -p -r1.77 crypto.c
--- crypto/crypto.c 15 Sep 2016 02:00:17 -  1.77
+++ crypto/crypto.c 16 Sep 2016 17:40:04 -
@@ -254,7 +254,7 @@ crypto_get_driverid(u_int8_t flags)
return -1;
}
 
-   bcopy(crypto_drivers, newdrv,
+   memcpy(newdrv, crypto_drivers,
crypto_drivers_num * sizeof(struct cryptocap));
bzero([crypto_drivers_num],
crypto_drivers_num * sizeof(struct cryptocap));
Index: crypto/gmac.c
===
RCS file: /cvs/src/sys/crypto/gmac.c,v
retrieving revision 1.8
diff -u -p -r1.8 gmac.c
--- crypto/gmac.c   7 Nov 2015 17:46:49 -   1.8
+++ crypto/gmac.c   16 Sep 2016 17:40:04 -
@@ -144,7 +144,7 @@ AES_GMAC_Update(void *xctx, const uint8_
(*ghash_update)(>ghash, (uint8_t *)data,
len - plen);
if (plen) {
-   bcopy((uint8_t *)data + (len - plen), (uint8_t *)blk,
+   memcpy((uint8_t *)blk, (uint8_t *)data + (len - plen),
plen);
(*ghash_update)(>ghash, (uint8_t *)blk,
GMAC_BLOCK_LEN);
Index: crypto/hmac.c
===
RCS file: /cvs/src/sys/crypto/hmac.c,v
retrieving revision 1.3
diff -u -p -r1.3 hmac.c
--- crypto/hmac.c   11 Jan 2011 15:42:05 -  1.3
+++ crypto/hmac.c   16 Sep 2016 17:40:04 -
@@ -46,7 +46,7 @@ HMAC_MD5_Init(HMAC_MD5_CTX *ctx, const u
}
 
bzero(k_ipad, MD5_BLOCK_LENGTH);
-   bcopy(ctx->key, k_ipad, ctx->key_len);
+   memcpy(k_ipad, ctx->key, ctx->key_len);
for (i = 0; i < MD5_BLOCK_LENGTH; i++)
k_ipad[i] ^= 0x36;
 
@@ -71,7 +71,7 @@ HMAC_MD5_Final(u_int8_t digest[MD5_DIGES
MD5Final(digest, >ctx);
 
bzero(k_opad, MD5_BLOCK_LENGTH);
-   bcopy(ctx->key, k_opad, ctx->key_len);
+   memcpy(k_opad, ctx->key, ctx->key_len);
for (i = 0; i < MD5_BLOCK_LENGTH; i++)
k_opad[i] ^= 0x5c;
 
@@ -100,7 +100,7 @@ HMAC_SHA1_Init(HMAC_SHA1_CTX *ctx, const
}
 
bzero(k_ipad, SHA1_BLOCK_LENGTH);
-   bcopy(ctx->key, k_ipad, ctx->key_len);
+   memcpy(k_ipad, ctx->key, ctx->key_len);
for (i = 0; i < SHA1_BLOCK_LENGTH; i++)
k_ipad[i] ^= 0x36;
 
@@ -125,7 +125,7 @@ HMAC_SHA1_Final(u_int8_t digest[SHA1_DIG
SHA1Final(digest, >ctx);
 
bzero(k_opad, SHA1_BLOCK_LENGTH);
-   bcopy(ctx->key, k_opad, ctx->key_len);
+   memcpy(k_opad, ctx->key, ctx->key_len);
for (i = 0; i < SHA1_BLOCK_LENGTH; i++)
k_opad[i] ^= 0x5c;
 
@@ -154,7 +154,7 @@ HMAC_SHA256_Init(HMAC_SHA256_CTX *ctx, c
}
 
bzero(k_ipad, SHA256_BLOCK_LENGTH);
-   bcopy(ctx->key, k_ipad, ctx->key_len);
+   memcpy(k_ipad, ctx->key, ctx->key_len);
for (i = 0; i < SHA256_BLOCK_LENGTH; i++)
k_ipad[i] ^= 0x36;
 
@@ -179,7 +179,7 @@ HMAC_SHA256_Final(u_int8_t digest[SHA256
SHA256Final(digest, >ctx);
 
bzero(k_opad, SHA256_BLOCK_LENGTH);
-   bcopy(ctx->key, k_opad, ctx->key_len);
+   memcpy(k_opad, ctx->key, ctx->key_len);
for (i = 0; i < SHA256_BLOCK_LENGTH; i++)
k_opad[i] ^= 0x5c;
 
Index: crypto/xform.c
===
RCS file: /cvs/src/sys/crypto/xform.c,v
retrieving revision 1.54
diff -u -p -r1.54 xform.c
--- crypto/xform.c  10 Dec 2015 21:00:51 -  1.54
+++ crypto/xform.c  16 Sep 2016 17:40:04 -
@@ -491,7 +491,7 @@ aes_xts_reinit(caddr_t key, u_int8_t *iv
 * Prepare tweak as E_k2(IV). IV is specified as LE representation
 * of a 64-bit block number which we allow to be passed in directly.
 */
-   bcopy(iv, , AES_XTS_IVSIZE);
+   memcpy(, iv, AES_XTS_IVSIZE);
for (i = 0; i < AES_XTS_IVSIZE; i++) {
ctx->tweak[i] = blocknum & 0xff;
blocknum >>= 8;

netinet bcopy->memcpy

[David Hill]

Hello -
Here is another diff to switch some bcopy's to memcpy's.

Most bcopy's are on freshly alloc'd memory.

for netinet/

Index: ip_ah.c
===
RCS file: /cvs/src/sys/netinet/ip_ah.c,v
retrieving revision 1.122
diff -u -p -r1.122 ip_ah.c
--- ip_ah.c 13 Sep 2016 19:56:55 -  1.122
+++ ip_ah.c 15 Sep 2016 16:14:33 -
@@ -151,7 +151,7 @@ ah_init(struct tdb *tdbp, struct xformsw
tdbp->tdb_amxkeylen = ii->ii_authkeylen;
tdbp->tdb_amxkey = malloc(tdbp->tdb_amxkeylen, M_XDATA, M_WAITOK);
 
-   bcopy(ii->ii_authkey, tdbp->tdb_amxkey, tdbp->tdb_amxkeylen);
+   memcpy(tdbp->tdb_amxkey, ii->ii_authkey, tdbp->tdb_amxkeylen);
 
/* Initialize crypto session. */
memset(, 0, sizeof(cria));
@@ -312,7 +312,7 @@ ah_massage_headers(struct mbuf **m0, int
 
/* Zeroize all other options. */
count = ptr[off + 1];
-   bcopy(ipseczeroes, ptr, count);
+   memcpy(ptr, ipseczeroes, count);
off += count;
break;
}
@@ -422,7 +422,7 @@ ah_massage_headers(struct mbuf **m0, int
 
/* If mutable option, zeroize. */
if (ptr[count] & IP6OPT_MUTABLE)
-   bcopy(ipseczeroes, ptr + count,
+   memcpy(ptr + count, ipseczeroes,
ptr[count + 1]);
 
count += ad;
@@ -642,7 +642,7 @@ ah_input(struct mbuf *m, struct tdb *tdb
 
if ((tdb->tdb_wnd > 0) && (tdb->tdb_flags & TDBF_ESN)) {
esn = htonl(esn);
-   bcopy(, crda->crd_esn, 4);
+   memcpy(crda->crd_esn, , 4);
crda->crd_flags |= CRD_F_ESN;
}
 
@@ -689,7 +689,7 @@ ah_input(struct mbuf *m, struct tdb *tdb
tc->tc_spi = tdb->tdb_spi;
tc->tc_proto = tdb->tdb_sproto;
tc->tc_rdomain = tdb->tdb_rdomain;
-   bcopy(>tdb_dst, >tc_dst, sizeof(union sockaddr_union));
+   memcpy(>tc_dst, >tdb_dst, sizeof(union sockaddr_union));
 
return crypto_dispatch(crp);
 }
@@ -1114,7 +1114,7 @@ ah_output(struct mbuf *m, struct tdb *td
u_int32_t esn;
 
esn = htonl((u_int32_t)(tdb->tdb_rpl >> 32));
-   bcopy(, crda->crd_esn, 4);
+   memcpy(crda->crd_esn, , 4);
crda->crd_flags |= CRD_F_ESN;
}
 
@@ -1138,9 +1138,8 @@ ah_output(struct mbuf *m, struct tdb *td
 */
switch (tdb->tdb_dst.sa.sa_family) {
case AF_INET:
-   bcopy(((caddr_t)(tc + 1)) +
-   offsetof(struct ip, ip_len),
-   (caddr_t) , sizeof(u_int16_t));
+   memcpy((caddr_t) , ((caddr_t)(tc + 1)) +
+   offsetof(struct ip, ip_len), sizeof(u_int16_t));
iplen = htons(ntohs(iplen) + rplen + ahx->authsize);
m_copyback(m, offsetof(struct ip, ip_len),
sizeof(u_int16_t), , M_NOWAIT);
@@ -1148,9 +1147,8 @@ ah_output(struct mbuf *m, struct tdb *td
 
 #ifdef INET6
case AF_INET6:
-   bcopy(((caddr_t)(tc + 1)) +
-   offsetof(struct ip6_hdr, ip6_plen),
-   (caddr_t) , sizeof(u_int16_t));
+   memcpy((caddr_t) , ((caddr_t)(tc + 1)) +
+   offsetof(struct ip6_hdr, ip6_plen), sizeof(u_int16_t));
iplen = htons(ntohs(iplen) + rplen + ahx->authsize);
m_copyback(m, offsetof(struct ip6_hdr, ip6_plen),
sizeof(u_int16_t), , M_NOWAIT);
@@ -1188,7 +1186,7 @@ ah_output(struct mbuf *m, struct tdb *td
tc->tc_spi = tdb->tdb_spi;
tc->tc_proto = tdb->tdb_sproto;
tc->tc_rdomain = tdb->tdb_rdomain;
-   bcopy(>tdb_dst, >tc_dst, sizeof(union sockaddr_union));
+   memcpy(>tc_dst, >tdb_dst, sizeof(union sockaddr_union));
 
return crypto_dispatch(crp);
 }
Index: ip_esp.c
===
RCS file: /cvs/src/sys/netinet/ip_esp.c,v
retrieving revision 1.140
diff -u -p -r1.140 ip_esp.c
--- ip_esp.c13 Sep 2016 19:56:55 -  1.140
+++ ip_esp.c15 Sep 2016 16:14:33 -
@@ -262,7 +262,7 @@ esp_init(struct tdb *tdbp, struct xforms
tdbp->tdb_emxkeylen = ii->ii_enckeylen;
tdbp->tdb_emxkey = malloc(tdbp->tdb_emxkeylen, M_XDATA,
M_WAITOK);
-   bcopy(ii->ii_enckey, tdbp->tdb_emxkey, tdbp->tdb_emxkeylen);
+   memcpy(tdbp->tdb_emxkey, ii->ii_enckey, tdbp->tdb_emxkeylen);
 
memset(, 0, sizeof(crie));
 
@@ -283,7 +283,7 @@ esp_init(struct tdb *tdbp, struct xforms
tdbp->tdb_amxkeylen = 

Re: netinet6 free() diff

[David Hill]

Hello -

Regenerated diff against -current. 

Index: netinet6/frag6.c
===
RCS file: /cvs/src/sys/netinet6/frag6.c,v
retrieving revision 1.67
diff -u -p -r1.67 frag6.c
--- netinet6/frag6.c7 Mar 2016 18:44:00 -   1.67
+++ netinet6/frag6.c19 Aug 2016 12:46:40 -
@@ -303,7 +303,7 @@ frag6_input(struct mbuf **mp, int *offp,
 
/* dequeue the fragment. */
LIST_REMOVE(af6, ip6af_list);
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
 
/* adjust pointer. */
ip6err = mtod(merr, struct ip6_hdr *);
@@ -348,14 +348,14 @@ frag6_input(struct mbuf **mp, int *offp,
ecn0 = (ntohl(af6->ip6af_flow) >> 20) & IPTOS_ECN_MASK;
if (ecn == IPTOS_ECN_CE) {
if (ecn0 == IPTOS_ECN_NOTECT) {
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto dropfrag;
}
if (ecn0 != IPTOS_ECN_CE)
af6->ip6af_flow |= htonl(IPTOS_ECN_CE << 20);
}
if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto dropfrag;
}
 
@@ -384,7 +384,7 @@ frag6_input(struct mbuf **mp, int *offp,
i,
inet_ntop(AF_INET6, >ip6q_src, ip, sizeof(ip)));
 #endif
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto flushfrags;
}
}
@@ -398,7 +398,7 @@ frag6_input(struct mbuf **mp, int *offp,
i,
inet_ntop(AF_INET6, >ip6q_src, ip, sizeof(ip)));
 #endif
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto flushfrags;
}
}
@@ -449,12 +449,12 @@ frag6_input(struct mbuf **mp, int *offp,
t = t->m_next;
t->m_next = IP6_REASS_MBUF(af6);
m_adj(t->m_next, af6->ip6af_offset);
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
}
 
/* adjust offset to point where the original next header starts */
offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
ip6->ip6_src = q6->ip6q_src;
@@ -465,7 +465,7 @@ frag6_input(struct mbuf **mp, int *offp,
if (frag6_deletefraghdr(m, offset) != 0) {
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
goto dropfrag;
}
@@ -480,7 +480,7 @@ frag6_input(struct mbuf **mp, int *offp,
 
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
 
if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
@@ -506,12 +506,12 @@ frag6_input(struct mbuf **mp, int *offp,
while ((af6 = LIST_FIRST(>ip6q_asfrag)) != NULL) {
LIST_REMOVE(af6, ip6af_list);
m_freem(IP6_REASS_MBUF(af6));
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
}
ip6stat.ip6s_fragdropped += q6->ip6q_nfrag;
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
 
  dropfrag:
@@ -579,11 +579,11 @@ frag6_freef(struct ip6q *q6)
ICMP6_TIME_EXCEED_REASSEMBLY, 0);
} else
m_freem(m);
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
}
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
 }
 
Index: netinet6/in6.c
===
RCS file: /cvs/src/sys/netinet6/in6.c,v
retrieving revision 1.190
diff -u -p -r1.190 in6.c
--- netinet6/in6.c  8 Aug 2016 13:09:36 -   1.190
+++ netinet6/in6.c  19 Aug 2016 12:46:41 -
@@ -1917,5 +1917,5 @@ in6_domifdetach(struct ifnet *ifp, void 

netinet6 free() diff

[David Hill]

Hello -

Here is a diff to add free sizes to netinet6/

Index: netinet6/frag6.c
===
RCS file: /cvs/src/sys/netinet6/frag6.c,v
retrieving revision 1.67
diff -u -p -r1.67 frag6.c
--- netinet6/frag6.c7 Mar 2016 18:44:00 -   1.67
+++ netinet6/frag6.c30 Jul 2016 17:17:14 -
@@ -303,7 +303,7 @@ frag6_input(struct mbuf **mp, int *offp,
 
/* dequeue the fragment. */
LIST_REMOVE(af6, ip6af_list);
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
 
/* adjust pointer. */
ip6err = mtod(merr, struct ip6_hdr *);
@@ -348,14 +348,14 @@ frag6_input(struct mbuf **mp, int *offp,
ecn0 = (ntohl(af6->ip6af_flow) >> 20) & IPTOS_ECN_MASK;
if (ecn == IPTOS_ECN_CE) {
if (ecn0 == IPTOS_ECN_NOTECT) {
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto dropfrag;
}
if (ecn0 != IPTOS_ECN_CE)
af6->ip6af_flow |= htonl(IPTOS_ECN_CE << 20);
}
if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto dropfrag;
}
 
@@ -384,7 +384,7 @@ frag6_input(struct mbuf **mp, int *offp,
i,
inet_ntop(AF_INET6, >ip6q_src, ip, sizeof(ip)));
 #endif
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto flushfrags;
}
}
@@ -398,7 +398,7 @@ frag6_input(struct mbuf **mp, int *offp,
i,
inet_ntop(AF_INET6, >ip6q_src, ip, sizeof(ip)));
 #endif
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
goto flushfrags;
}
}
@@ -449,12 +449,12 @@ frag6_input(struct mbuf **mp, int *offp,
t = t->m_next;
t->m_next = IP6_REASS_MBUF(af6);
m_adj(t->m_next, af6->ip6af_offset);
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
}
 
/* adjust offset to point where the original next header starts */
offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
-   free(ip6af, M_FTABLE, 0);
+   free(ip6af, M_FTABLE, sizeof(*ip6af));
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
ip6->ip6_src = q6->ip6q_src;
@@ -465,7 +465,7 @@ frag6_input(struct mbuf **mp, int *offp,
if (frag6_deletefraghdr(m, offset) != 0) {
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
goto dropfrag;
}
@@ -480,7 +480,7 @@ frag6_input(struct mbuf **mp, int *offp,
 
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
 
if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
@@ -506,12 +506,12 @@ frag6_input(struct mbuf **mp, int *offp,
while ((af6 = LIST_FIRST(>ip6q_asfrag)) != NULL) {
LIST_REMOVE(af6, ip6af_list);
m_freem(IP6_REASS_MBUF(af6));
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
}
ip6stat.ip6s_fragdropped += q6->ip6q_nfrag;
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
 
  dropfrag:
@@ -579,11 +579,11 @@ frag6_freef(struct ip6q *q6)
ICMP6_TIME_EXCEED_REASSEMBLY, 0);
} else
m_freem(m);
-   free(af6, M_FTABLE, 0);
+   free(af6, M_FTABLE, sizeof(*af6));
}
TAILQ_REMOVE(_queue, q6, ip6q_queue);
frag6_nfrags -= q6->ip6q_nfrag;
-   free(q6, M_FTABLE, 0);
+   free(q6, M_FTABLE, sizeof(*q6));
frag6_nfragpackets--;
 }
 
Index: netinet6/in6.c
===
RCS file: /cvs/src/sys/netinet6/in6.c,v
retrieving revision 1.189
diff -u -p -r1.189 in6.c
--- netinet6/in6.c  13 Jul 2016 08:40:46 -  1.189
+++ netinet6/in6.c  30 Jul 2016 17:17:15 -
@@ -1915,5 +1915,5 @@ in6_domifdetach(struct ifnet *ifp, void 

Re: nfs_vnops: nfs_lookitup - uninitialized var

[David Hill]

Any thoughts on this?

On Mon, Jun 06, 2016 at 01:36:05PM -0400, David Hill wrote:
> Hello -
> 
> Clang reports a possible user of an uninitalized variable in
> nfs_vnops.c line 2605.
> 
> attrflag is uninitialized when calling nfsm_postop_attr(), which is a
> macro that only sets attrflag if (info.nmi_mrep != NULL).  I am not sure
> if that is possible, but maybe a KASSERT(info.nmi_mrep != NULL) before
> nfsm_postop_attr() would catch a bug...
> 
> Or just silence it with:
> 
> Index: nfs_vnops.c
> ===
> RCS file: /cvs/src/sys/nfs/nfs_vnops.c,v
> retrieving revision 1.169
> diff -u -p -r1.169 nfs_vnops.c
> --- nfs_vnops.c   29 Apr 2016 14:40:36 -  1.169
> +++ nfs_vnops.c   6 Jun 2016 17:34:24 -
> @@ -2562,7 +2562,7 @@ nfs_lookitup(struct vnode *dvp, char *na
>   struct vnode *newvp = NULL;
>   struct nfsnode *np, *dnp = VTONFS(dvp);
>   caddr_t cp2;
> - int error = 0, fhlen, attrflag;
> + int error = 0, fhlen, attrflag = 0;
>   nfsfh_t *nfhp;
>  
>   info.nmi_v3 = NFS_ISV3(dvp);
> 

nfs_vnops: nfs_lookitup - uninitialized var

[David Hill]

Hello -

Clang reports a possible user of an uninitalized variable in
nfs_vnops.c line 2605.

attrflag is uninitialized when calling nfsm_postop_attr(), which is a
macro that only sets attrflag if (info.nmi_mrep != NULL).  I am not sure
if that is possible, but maybe a KASSERT(info.nmi_mrep != NULL) before
nfsm_postop_attr() would catch a bug...

Or just silence it with:

Index: nfs_vnops.c
===
RCS file: /cvs/src/sys/nfs/nfs_vnops.c,v
retrieving revision 1.169
diff -u -p -r1.169 nfs_vnops.c
--- nfs_vnops.c 29 Apr 2016 14:40:36 -  1.169
+++ nfs_vnops.c 6 Jun 2016 17:34:24 -
@@ -2562,7 +2562,7 @@ nfs_lookitup(struct vnode *dvp, char *na
struct vnode *newvp = NULL;
struct nfsnode *np, *dnp = VTONFS(dvp);
caddr_t cp2;
-   int error = 0, fhlen, attrflag;
+   int error = 0, fhlen, attrflag = 0;
nfsfh_t *nfhp;
 
info.nmi_v3 = NFS_ISV3(dvp);

pcmcia unused function diff

[David Hill]

Hello -

Remove unused functions. 

Index: dev/pcmcia/pcmcia.c
===
RCS file: /cvs/src/sys/dev/pcmcia/pcmcia.c,v
retrieving revision 1.46
diff -u -p -r1.46 pcmcia.c
--- dev/pcmcia/pcmcia.c 8 Oct 2012 21:47:50 -   1.46
+++ dev/pcmcia/pcmcia.c 2 Jun 2016 16:11:18 -
@@ -58,13 +58,7 @@ void pcmcia_attach(struct device *, stru
 intpcmcia_activate(struct device *, int);
 intpcmcia_print(void *, const char *);
 void   pcmcia_card_detach_notify(struct device *, void *);
-
-static inline void pcmcia_socket_enable(pcmcia_chipset_tag_t,
-pcmcia_chipset_handle_t *);
-static inline void pcmcia_socket_disable(pcmcia_chipset_tag_t,
- pcmcia_chipset_handle_t *);
-
-int pcmcia_card_intr(void *);
+intpcmcia_card_intr(void *);
 
 struct cfdriver pcmcia_cd = {
NULL, "pcmcia", DV_DULL
@@ -409,20 +403,6 @@ pcmcia_function_init(pf, cfe)
 
/* Remember which configuration entry we are using. */
pf->cfe = cfe;
-}
-
-static inline void pcmcia_socket_enable(pct, pch)
- pcmcia_chipset_tag_t pct;
- pcmcia_chipset_handle_t *pch;
-{
-   pcmcia_chip_socket_enable(pct, pch);
-}
-
-static inline void pcmcia_socket_disable(pct, pch)
- pcmcia_chipset_tag_t pct;
- pcmcia_chipset_handle_t *pch;
-{
-   pcmcia_chip_socket_disable(pct, pch);
 }
 
 /* Enable a PCMCIA function */

Re: xhci(4) vs BIOS

[David Hill]

On Tue, May 31, 2016 at 06:43:16PM +0200, Martin Pieuchot wrote:
> I finally got my hands on a DELL Optiplex 3020.  The BIOS on this
> machines uses the xHCI controller to handle the keyboard and mouse.
> 
> Turns out that I badly copy/pasted the ehci(4) code to take over the
> HC from the BIOS.  With the diff below I can now route all my ports
> to xhci(4) and the machine no longer hang at boot.
> 
> ok?
> 

Fixes my Dell Latitude.

Thanks!
David

Re: kernel panic - rtable_match

[David Hill]

Great!  I appreciate the commit hint.

Thanks

On Fri, May 27, 2016 at 10:46:32PM +0100, Stuart Henderson wrote:
> On 2016/05/27 17:29, David Hill wrote:
> > Hello -
> > 
> > Running openbsd current via cvs from today..
> > 
> > My machine keeps panic'ing with a uvm_fault in rtable_match.
> > It locks my whole machine up, so I am unable to type anything.  Perhaps
> > someone can provide me some debugging printf's?
> 
> I sent mpi some more details when I hit it earlier. Backtrace is here:
> 
> rtable_match() at rtable_match+0x25
> rt_match() at rt_match+0x7b
> _rtalloc() at _rtalloc+0x1f
> pf_refragment6() at pf_refragment6+0x1f4
> pf_test() at pf_test+0x9ba
> ip6_forward() at ip6_forward+0x20f 
> ip6_input() at ip6_input+0xad9
> ip6intr() at ip6intr+0x18
> if_netisr() at if_netisr+0xdc
> taskq_thread() at taskq_thread+0x6c
> 
> I haven't seen it again after backing out this commit:
> 
> http://anoncvs.spacehopper.org/openbsd-sys/commit/?id=ffcc809d2ffdc00af088aa5285957d2472577a9e
> 

kernel panic - rtable_match

[David Hill]

Hello -

Running openbsd current via cvs from today..

My machine keeps panic'ing with a uvm_fault in rtable_match.
It locks my whole machine up, so I am unable to type anything.  Perhaps
someone can provide me some debugging printf's?

uvm_fault(0x81945e00, 0x1, 0, 1) -> e
kernel: page fault trap, code=0
Stopped at rtable_match+0x25:  movzbl 0x1(%rsi),%esi
ddb{4}>

0x790 + 0x25 = 7b5

0790 :
 790:   55  push   %rbp
 791:   48 89 e5mov%rsp,%rbp
 794:   41 57   push   %r15
 796:   41 56   push   %r14
 798:   41 55   push   %r13
 79a:   49 89 f5mov%rsi,%r13
 79d:   41 54   push   %r12
 79f:   53  push   %rbx
 7a0:   48 83 ec 28 sub$0x28,%rsp
 7a4:   48 89 55 b8 mov%rdx,0xffb8(%rbp)
 7a8:   48 8b 05 00 00 00 00mov0(%rip),%rax# 7af 

 7af:   48 89 45 d0 mov%rax,0xffd0(%rbp)
 7b3:   31 c0   xor%eax,%eax
-->  7b5:   0f b6 76 01 movzbl 0x1(%rsi),%esi
 7b9:   e8 00 00 00 00  callq  7be 
 7be:   48 85 c0test   %rax,%rax
 7c1:   48 89 c3mov%rax,%rbx
 7c4:   48 c7 45 c0 00 00 00movq   $0x0,0xffc0(%rbp)
 7cb:   00 
 7cc:   0f 84 13 01 00 00   je 8e5 
 7d2:   0f b6 40 2a movzbl 0x2a(%rax),%eax
 7d6:   4e 8d 24 28 lea(%rax,%r13,1),%r12
 7da:   e8 00 00 00 00  callq  7df 
 7df:   4c 89 e6mov%r12,%rsi
 7e2:   48 89 dfmov%rbx,%rdi
 7e5:   e8 00 00 00 00  callq  7ea 
 7ea:   48 85 c0test   %rax,%rax
 7ed:   49 89 c7mov%rax,%r15
 7f0:   48 c7 45 c0 00 00 00movq   $0x0,0xffc0(%rbp)
 7f7:   00 
 7f8:   0f 84 e2 00 00 00   je 8e0 
 7fe:   4c 8d 75 c8 lea0xffc8(%rbp),%r14
 802:   48 89 c6mov%rax,%rsi
 805:   4c 89 f7mov%r14,%rdi
 808:   e8 00 00 00 00  callq  80d 
 80d:   48 89 c7mov%rax,%rdi
 810:   48 89 45 c0 mov%rax,0xffc0(%rbp)
 814:   e8 00 00 00 00  callq  819 
 819:   4c 89 f7mov%r14,%rdi
 81c:   e8 00 00 00 00  callq  821 
 821:   48 8b 55 b8 mov0xffb8(%rbp),%rdx
 825:   48 8b 7d c0 mov0xffc0(%rbp),%rdi
 829:   4c 89 eemov%r13,%rsi
 82c:   e8 00 00 00 00  callq  831 
 831:   83 f8 ffcmp$0x,%eax
 834:   41 89 c4mov%eax,%r12d
 837:   0f 84 a3 00 00 00   je 8e0 
 83d:   3d ff ff 00 00  cmp$0x,%eax
 842:   0f 8f fd 00 00 00   jg 945 
 848:   4c 89 femov%r15,%rsi
 84b:   4c 89 f7mov%r14,%rdi
 84e:   31 db   xor%ebx,%ebx
 850:   e8 00 00 00 00  callq  855 
 855:   48 85 c0test   %rax,%rax
 858:   74 2d   je 887 
 85a:   66  data16
 85b:   66  data16
 85c:   90  nop
 85d:   66  data16
 85e:   66  data16
 85f:   90  nop
 860:   48 8b 4d c0 mov0xffc0(%rbp),%rcx
 864:   48 89 c2mov%rax,%rdx
 867:   0f b6 40 6a movzbl 0x6a(%rax),%eax
 86b:   48 8d 72 08 lea0x8(%rdx),%rsi
 86f:   4c 89 f7mov%r14,%rdi
 872:   3a 41 6acmp0x6a(%rcx),%al
 875:   0f 94 c0sete   %al
 878:   0f b6 c0movzbl %al,%eax
 87b:   01 c3   add%eax,%ebx
 87d:   e8 00 00 00 00  callq  882 
 882:   48 85 c0test   %rax,%rax
 885:   75 d9   jne860 
 887:   4c 89 f7mov%r14,%rdi
 88a:   e8 00 00 00 00  callq  

Re: [diff] xhci intel port route freeze

[David Hill]

Ugh, not enough testing.  Breaks again on resume.  Ignore this diff.
Too excited about being able to charge my phone with my laptop :)

- David

[diff] xhci intel port route freeze

[David Hill]

Hello -

Looking at the FreeBSD code, their xhci_pci_port_route function is
called on activation and not on attach.

OpenBSD calles it in xhci_pci_attach and xhci_pci_activate.

If I remove it from attach, all my usb devices attach and no longer
lock the machine.
ugen1 at uhub4 port 2 "Apple Inc. iPhone" rev 2.00/7.01 addr 3

FreeBSD sets the port route function to use if intel in attach but does
not call it:
https://github.com/freebsd/freebsd/blob/5badcaaa0f325d543c4a79c30c6c91b0e88163fd/sys/dev/usb/controller/xhci_pci.c#L228

FreeBSD calls it in activate:
https://github.com/freebsd/freebsd/blob/0f2211c30ded7871b441bfaa1647843f1b4fd0c2/sys/dev/usb/controller/xhci.c#L485

OpenBSD might be missing:
https://github.com/freebsd/freebsd/blob/0f2211c30ded7871b441bfaa1647843f1b4fd0c2/sys/dev/usb/controller/xhci.c#L1439-L1445

Index: pci/xhci_pci.c
===
RCS file: /cvs/src/sys/dev/pci/xhci_pci.c,v
retrieving revision 1.7
diff -u -p -r1.7 xhci_pci.c
--- pci/xhci_pci.c  2 Nov 2015 14:53:10 -   1.7
+++ pci/xhci_pci.c  6 Feb 2016 16:23:06 -
@@ -192,9 +192,6 @@ xhci_pci_attach(struct device *parent, s
goto disestablish_ret;
}
 
-   if (PCI_VENDOR(psc->sc_id) == PCI_VENDOR_INTEL)
-   xhci_pci_port_route(psc);
-
/* Attach usb device. */
config_found(self, >sc.sc_bus, usbctlprint);
 

[diff] IFQ has its own protection

[David Hill]

Hello -

IFQ has its own protection, so no need for splnet/splx.

if_gif.c - use for (;;) to match the other drivers.


Index: net/if_gif.c
===
RCS file: /cvs/src/sys/net/if_gif.c,v
retrieving revision 1.83
diff -u -p -r1.83 if_gif.c
--- net/if_gif.c5 Dec 2015 10:07:55 -   1.83
+++ net/if_gif.c14 Jan 2016 00:26:10 -
@@ -164,13 +164,9 @@ gif_start(struct ifnet *ifp)
 {
struct gif_softc *sc = (struct gif_softc*)ifp;
struct mbuf *m;
-   int s;
 
-   while (1) {
-   s = splnet();
+   for (;;) {
IFQ_DEQUEUE(>if_snd, m);
-   splx(s);
-
if (m == NULL)
break;
 
Index: net/if_mpe.c
===
RCS file: /cvs/src/sys/net/if_mpe.c,v
retrieving revision 1.52
diff -u -p -r1.52 if_mpe.c
--- net/if_mpe.c5 Dec 2015 10:07:55 -   1.52
+++ net/if_mpe.c14 Jan 2016 00:26:10 -
@@ -143,16 +143,12 @@ mpestart(struct ifnet *ifp0)
 {
struct mbuf *m;
struct sockaddr *sa = (struct sockaddr *)
-   int  s;
sa_family_t  af;
struct rtentry  *rt;
struct ifnet*ifp;
 
for (;;) {
-   s = splnet();
IFQ_DEQUEUE(>if_snd, m);
-   splx(s);
-
if (m == NULL)
return;
 
Index: net/if_pppx.c
===
RCS file: /cvs/src/sys/net/if_pppx.c,v
retrieving revision 1.48
diff -u -p -r1.48 if_pppx.c
--- net/if_pppx.c   5 Dec 2015 16:09:09 -   1.48
+++ net/if_pppx.c   14 Jan 2016 00:26:10 -
@@ -995,15 +995,13 @@ pppx_if_start(struct ifnet *ifp)
 {
struct pppx_if *pxi = (struct pppx_if *)ifp->if_softc;
struct mbuf *m;
-   int proto, s;
+   int proto;
 
if (!ISSET(ifp->if_flags, IFF_RUNNING))
return;
 
for (;;) {
-   s = splnet();
IFQ_DEQUEUE(>if_snd, m);
-   splx(s);
 
if (m == NULL)
break;
Index: net/if_vxlan.c
===
RCS file: /cvs/src/sys/net/if_vxlan.c,v
retrieving revision 1.35
diff -u -p -r1.35 if_vxlan.c
--- net/if_vxlan.c  5 Dec 2015 10:07:55 -   1.35
+++ net/if_vxlan.c  14 Jan 2016 00:26:10 -
@@ -275,15 +275,12 @@ void
 vxlanstart(struct ifnet *ifp)
 {
struct mbuf *m;
-   int  s;
 
for (;;) {
-   s = splnet();
IFQ_DEQUEUE(>if_snd, m);
-   splx(s);
-
if (m == NULL)
return;
+
ifp->if_opackets++;
 
 #if NBPFILTER > 0

uvm_io: uiomovei -> uiomove

[David Hill]

Hello

sz is type vsize_t, and vsize_t and vaddr_t are unsigned long on
every arch, so I believe this can safely be converted.
   
vaddr_t baseva, endva, pageoffset, kva;
vsize_t chunksz, togo, sz; 

- David

Index: uvm/uvm_io.c
===
RCS file: /cvs/src/sys/uvm/uvm_io.c,v
retrieving revision 1.25
diff -u -p -r1.25 uvm_io.c
--- uvm/uvm_io.c14 Mar 2015 03:38:53 -  1.25
+++ uvm/uvm_io.c31 Dec 2015 21:46:14 -
@@ -109,7 +109,7 @@ uvm_io(vm_map_t map, struct uio *uio, in
sz = chunksz - pageoffset;
if (sz > togo)
sz = togo;
-   error = uiomovei((caddr_t) (kva + pageoffset), sz, uio);
+   error = uiomove((caddr_t) (kva + pageoffset), sz, uio);
togo -= sz;
baseva += chunksz;
 

Re: initial 11n support for iwn (n, not m)

[David Hill]

On Fri, Dec 18, 2015 at 08:52:07PM +, Stuart Henderson wrote:
> On 2015/12/17 19:43, Frank Groeneveld wrote:
> > iwn0: fatal firmware error
> 
> So this firmware crash...
> 
> > firmware error log:
> >   error type  = "UNKNOWN" (0x14E2)
> >   program counter = 0x00018100
> >   source line = 0x0216
> >   error data  = 0x0028
> >   branch link = 0x000180E2000180E2
> >   interrupt link  = 0xDBEA
> >   time= 3006541902
> 
> I'm seeing the same crash when I associate to a 2GHz channel where
> the AP has set a flag in beacons to warn stations that there are
> non-11n devices associated.
> 
> Around line 4900 in sys/dev/pci/if_iwn.c you have this:
> 
> case IEEE80211_HTPROT_NONE:
> break;
> case IEEE80211_HTPROT_NONMEMBER:
> case IEEE80211_HTPROT_NONHT_MIXED:
> sc->rxon.flags |=
> htole32(IWN_RXON_HT_RXON_HT_MODE_MIXED);
> break;
> 
> Try this as I suspect it might fix your crash (unless Stefan
> comes up with a better change to try first :)
> 
>   case IEEE80211_HTPROT_NONE:
>   break;
>   case IEEE80211_HTPROT_NONMEMBER:
>   case IEEE80211_HTPROT_NONHT_MIXED:
>   if (IEEE80211_IS_CHAN_5GHZ(ni->ni_chan))
>   sc->rxon.flags |=
>   htole32(IWN_RXON_HT_RXON_HT_MODE_MIXED);
>   break;
>

I had the firmware bug on Centrino Ultimate-N 6300:
iwn0: fatal firmware error
firmware error log:
  error type  = "SYSASSERT" (0x0005)
  program counter = 0x00022278
  source line = 0x0218
  error data  = 0x0218000B
  branch link = 0x0002225800022258
  interrupt link  = 0x1532
  time= 2127856977


With sthen@'s patch I can associate, dhcp, and use net.

Nice find!

- David

Re: initial 11n support for iwn (n, not m)

[David Hill]

On Wed, Dec 16, 2015 at 05:15:27PM +0100, Stefan Sperling wrote:
> On Wed, Dec 16, 2015 at 10:14:49AM -0500, David Hill wrote:
> > Hi Stefan -
> > 
> > Thanks for the 11n work!
> > 
> > Unfortunately, your diff breaks iwn on my machine.
> > 
> > iwn0 at pci2 dev 0 function 0 "Intel Centrino Ultimate-N 6300" rev 0x35:
> > msi, MIMO 3T3R, MoW, 
> > 
> > It spews over and over:
> > 
> > iwn0: fatal firmware error
> > firmware error log:
> >   error type  = "SYSASSERT" (0x0005)
> >   program counter = 0x00022278
> >   source line = 0x0218
> >   error data  = 0x0218000B
> >   branch link = 0x0002225800022258
> >   interrupt link  = 0x1532
> >   time= 2127856977
> > driver status:
> >   tx ring  0: qid=0  cur=4   queued=0  
> >   tx ring  1: qid=1  cur=0   queued=0  
> >   tx ring  2: qid=2  cur=0   queued=0  
> >   tx ring  3: qid=3  cur=0   queued=0  
> >   tx ring  4: qid=4  cur=41  queued=0  
> >   tx ring  5: qid=5  cur=0   queued=0  
> >   tx ring  6: qid=6  cur=0   queued=0  
> >   tx ring  7: qid=7  cur=0   queued=0  
> >   tx ring  8: qid=8  cur=0   queued=0  
> >   tx ring  9: qid=9  cur=0   queued=0  
> >   tx ring 10: qid=10 cur=0   queued=0  
> >   tx ring 11: qid=11 cur=0   queued=0  
> >   tx ring 12: qid=12 cur=0   queued=0  
> >   tx ring 13: qid=13 cur=0   queued=0  
> >   tx ring 14: qid=14 cur=0   queued=0  
> >   tx ring 15: qid=15 cur=0   queued=0  
> >   tx ring 16: qid=16 cur=0   queued=0  
> >   tx ring 17: qid=17 cur=0   queued=0  
> >   tx ring 18: qid=18 cur=0   queued=0  
> >   tx ring 19: qid=19 cur=0   queued=0  
> >   rx ring: cur=14
> >   802.11 state 4
> > 
> 
> Thanks for testing!
> 
> I cannot do much based on the information provided.
> Could you recompile with IWM_DEBUG defined, and perhaps place a few
> additional printfs at strategic locations, to figure out which
> firmware command is last sent before the firmware crashes?
> That would help me a great deal.
> 
> If you don't want the firmware to be restarted over and over so it
> won't print these lines repeatedly, disabling the init_task which
> attempts to recover from firmware crashes might help:
> 
>   if (r1 & (IWN_INT_SW_ERR | IWN_INT_HW_ERR)) {
>   printf("%s: fatal firmware error\n", sc->sc_dev.dv_xname);
>   /* Dump firmware error log and stop. */
>   iwn_fatal_intr(sc);
>   iwn_stop(ifp, 1);
>   task_add(systq, >init_task); <-- remove this line
>   return 1;
>   }
>


Little more time to play, but not much. Will play more tonight.

I do not have open wifi to test with, wpakey required.

16:21:14.335908 802.11 flags=0<>: probe response,
caps=2061<ESS,PRIVACY,SHORT_PREAMBLE,SHORT_SLOTTIME>, ssid (wifi),
rates 1M 2M 5M 11M 6M 9M 12M 18M, ds (chan 11), country 'US ', erp 0x00,
rsn 0x010fac04010fac04010fac02, xrates 24M 36M 48M 54M,
htcaps=<20MHz,TXSTBC,RXSTBC 1 stream,A-MSDU 3839,A-MPDU max 65535,A-MPDU
spacing 8.00us,RxMCS 0x>, 

ifconfig iwn0 up works
ifconfig iwn0 scan works

but as soon as I use nwid/wpakey to associate, it bombs.

...

sending scan command nchan=24
scan finished nchan=24 status=1 chan=165
sending scan command nchan=13
scan finished nchan=13 status=1 chan=13
sending scan command nchan=24
scan finished nchan=24 status=1 chan=165
rxon chan 11 flags 40008025 cck f ofdm 15 
setting TX power
adding broadcast node
timing bintval=400, tstamp=5529864295299, init=306301
iwn_run: htprot = 3
rxon chan 11 flags 44008035 cck f ofdm 15
setting TX power
adding BSS node
setting link quality for node 0 
setting initial differential gains
sending request for statistics
iwn0: fatal firmware error
firmware error log:
  error type  = "SYSASSERT" (0x0005)
  program counter = 0x00022278
  source line = 0x0218
  error data  = 0x0218000B
  branch link = 0x0002225800022258
  interrupt link  = 0x1532
  time= 2241458206
driver status:
  tx ring  0: qid=0  cur=2   queued=0
  tx ring  1: qid=1  cur=0   queued=0
  tx ring  2: qid=2  cur=0   queued=0
  tx ring  3: qid=3  cur=0   queued=0
  tx ring  4: qid=4  cur=63  queued=0
  tx ring  5: qid=5  cur=0   queued=0
  tx ring  6: qid=6  cur=0   queued=0
  tx ring  7: qid=7  cur=0   queued=0
  tx ring  8: qid=8  cur=0   queued=0
  tx ring  9: qid=9  cur=0   queued=0
  tx ring 10: qid=10 cur=0   queued=0
  tx ring 11: qid=11 cur=0   queued=0
  tx ring 12: qid=12 cur=0   queued=0
  tx ring 13: qid=13 cur=0   queued=0
  tx ring 14: qid=14 cur=0   queued=0
  tx ring 15: qid=15 cur=0   queued=0
  tx ring 16: qid=16 cur=0   queued=0
  tx ring 17: qid=17 cur=0   queued=0
  tx ring 18: qid=18 cur=0   queued=0
  tx ring 19: qid=19 cur=0   queued=0
  rx ring: cur=20
  802.11 state 4

Re: initial 11n support for iwn (n, not m)

[David Hill]

Hi Stefan -

Thanks for the 11n work!

Unfortunately, your diff breaks iwn on my machine.

iwn0 at pci2 dev 0 function 0 "Intel Centrino Ultimate-N 6300" rev 0x35:
msi, MIMO 3T3R, MoW, 

It spews over and over:

iwn0: fatal firmware error
firmware error log:
  error type  = "SYSASSERT" (0x0005)
  program counter = 0x00022278
  source line = 0x0218
  error data  = 0x0218000B
  branch link = 0x0002225800022258
  interrupt link  = 0x1532
  time= 2127856977
driver status:
  tx ring  0: qid=0  cur=4   queued=0  
  tx ring  1: qid=1  cur=0   queued=0  
  tx ring  2: qid=2  cur=0   queued=0  
  tx ring  3: qid=3  cur=0   queued=0  
  tx ring  4: qid=4  cur=41  queued=0  
  tx ring  5: qid=5  cur=0   queued=0  
  tx ring  6: qid=6  cur=0   queued=0  
  tx ring  7: qid=7  cur=0   queued=0  
  tx ring  8: qid=8  cur=0   queued=0  
  tx ring  9: qid=9  cur=0   queued=0  
  tx ring 10: qid=10 cur=0   queued=0  
  tx ring 11: qid=11 cur=0   queued=0  
  tx ring 12: qid=12 cur=0   queued=0  
  tx ring 13: qid=13 cur=0   queued=0  
  tx ring 14: qid=14 cur=0   queued=0  
  tx ring 15: qid=15 cur=0   queued=0  
  tx ring 16: qid=16 cur=0   queued=0  
  tx ring 17: qid=17 cur=0   queued=0  
  tx ring 18: qid=18 cur=0   queued=0  
  tx ring 19: qid=19 cur=0   queued=0  
  rx ring: cur=14
  802.11 state 4

netinet6 - bcopy -> memcpy

[David Hill]

Hello -

Here are a few bcopy -> memcpy conversions for buffers that do not
overlap.

Index: sys/netinet6/ip6_forward.c
===
RCS file: /cvs/src/sys/netinet6/ip6_forward.c,v
retrieving revision 1.86
diff -u -p -r1.86 ip6_forward.c
--- sys/netinet6/ip6_forward.c  2 Nov 2015 07:22:28 -   1.86
+++ sys/netinet6/ip6_forward.c  2 Dec 2015 17:20:27 -
@@ -206,7 +206,7 @@ reroute:
}
 
/* We need to do IPsec */
-   bcopy(>tdb_dst, , sizeof(sdst));
+   memcpy(, >tdb_dst, sizeof(sdst));
sspi = tdb->tdb_spi;
sproto = tdb->tdb_sproto;
}
Index: sys/netinet6/ip6_input.c
===
RCS file: /cvs/src/sys/netinet6/ip6_input.c,v
retrieving revision 1.151
diff -u -p -r1.151 ip6_input.c
--- sys/netinet6/ip6_input.c11 Nov 2015 10:23:23 -  1.151
+++ sys/netinet6/ip6_input.c2 Dec 2015 17:20:28 -
@@ -804,7 +804,7 @@ ip6_process_hopopts(struct mbuf *m, u_in
return (-1);
}
optlen = IP6OPT_RTALERT_LEN;
-   bcopy((caddr_t)(opt + 2), (caddr_t)_val, 2);
+   memcpy(_val, opt + 2, sizeof(rtalert_val));
*rtalertp = ntohs(rtalert_val);
break;
case IP6OPT_JUMBO:
@@ -837,9 +837,9 @@ ip6_process_hopopts(struct mbuf *m, u_in
 
/*
 * We may see jumbolen in unaligned location, so
-* we'd need to perform bcopy().
+* we'd need to perform memcpy().
 */
-   bcopy(opt + 2, , sizeof(jumboplen));
+   memcpy(, opt + 2, sizeof(jumboplen));
jumboplen = (u_int32_t)htonl(jumboplen);
 
 #if 1
@@ -962,7 +962,7 @@ ip6_savecontrol(struct inpcb *in6p, stru
/* RFC 2292 sec. 5 */
if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
struct in6_pktinfo pi6;
-   bcopy(>ip6_dst, _addr, sizeof(struct in6_addr));
+   memcpy(_addr, >ip6_dst, sizeof(struct in6_addr));
if (IN6_IS_SCOPE_EMBED(_addr))
pi6.ipi6_addr.s6_addr16[1] = 0;
pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
Index: sys/netinet6/ip6_output.c
===
RCS file: /cvs/src/sys/netinet6/ip6_output.c,v
retrieving revision 1.201
diff -u -p -r1.201 ip6_output.c
--- sys/netinet6/ip6_output.c   2 Dec 2015 13:29:26 -   1.201
+++ sys/netinet6/ip6_output.c   2 Dec 2015 17:20:28 -
@@ -2015,7 +2015,7 @@ do {\
dst->type = malloc(hlen, M_IP6OPT, canwait);\
if (dst->type == NULL && canwait == M_NOWAIT)\
goto bad;\
-   bcopy(src->type, dst->type, hlen);\
+   memcpy(dst->type, src->type, hlen);\
}\
 } while (/*CONSTCOND*/ 0)
 
Index: sys/netinet6/raw_ip6.c
===
RCS file: /cvs/src/sys/netinet6/raw_ip6.c,v
retrieving revision 1.87
diff -u -p -r1.87 raw_ip6.c
--- sys/netinet6/raw_ip6.c  24 Nov 2015 13:37:16 -  1.87
+++ sys/netinet6/raw_ip6.c  2 Dec 2015 17:20:28 -
@@ -756,7 +756,7 @@ rip6_usrreq(struct socket *so, int req, 
bzero(, sizeof(tmp));
tmp.sin6_family = AF_INET6;
tmp.sin6_len = sizeof(struct sockaddr_in6);
-   bcopy(>inp_faddr6, _addr,
+   memcpy(_addr, >inp_faddr6,
sizeof(struct in6_addr));
dst = 
} else {

Re: netstat -P - no namelist

[David Hill]

On Fri, Nov 20, 2015 at 12:07:36PM -0500, David Hill wrote:
> Hello -
> 
> I used to use netstat -P to view PCB information.  Running -current, I
> get 'no namelist' now.  Any ideas? 
> 
> 0xff0397c4da50 tcp  0  0  127.0.0.1.33845127.0.0.1.18333  
>   ESTABLISHED
> 
> # netstat -P 0xff0397c4da50
> netstat: no namelist
> 
> Thanks,
> David 
>

I am not sure if this is a fix or a hint, but it does provide me with
the information I used to get.  Did something change that would cause
n_type to always be 0 (or unset?)

Index: main.c
===
RCS file: /cvs/src/usr.bin/netstat/main.c,v
retrieving revision 1.108
diff -u -p -r1.108 main.c
--- main.c  23 Oct 2015 08:18:57 -  1.108
+++ main.c  23 Nov 2015 15:50:57 -
@@ -331,7 +331,7 @@ main(int argc, char *argv[])
(need_nlist ? 0 : KVM_NO_FILES), buf)) == NULL)
errx(1, "kvm_openfiles: %s", buf);
 
-   if (need_nlist && (kvm_nlist(kvmd, nl) < 0 || nl[0].n_type == 0)) {
+   if (need_nlist && (kvm_nlist(kvmd, nl) < 0)) {
if (nlistf)
errx(1, "%s: no namelist", nlistf);
else

netstat -P - no namelist

[David Hill]

Hello -

I used to use netstat -P to view PCB information.  Running -current, I
get 'no namelist' now.  Any ideas? 

0xff0397c4da50 tcp  0  0  127.0.0.1.33845127.0.0.1.18333
ESTABLISHED

# netstat -P 0xff0397c4da50
netstat: no namelist

Thanks,
David 

tcpbench pledge

[David Hill]

Hello -

pledge starts after getopt because of setrtable. 

rpath needed incase -k (kvm_openfile) 
proc needed for drop_gid (setresgid)

I believe I've hit every code path.  More eyes are welcome.

Index: tcpbench.c
===
RCS file: /cvs/src/usr.bin/tcpbench/tcpbench.c,v
retrieving revision 1.46
diff -u -p -r1.46 tcpbench.c
--- tcpbench.c  9 Feb 2015 23:46:23 -   1.46
+++ tcpbench.c  10 Nov 2015 16:43:16 -
@@ -986,8 +986,6 @@ quit(int sig, short event, void *arg)
 int
 main(int argc, char **argv)
 {
-   extern int optind;
-   extern char *optarg;
struct timeval tv;
unsigned int secs, rtable;
 
@@ -1108,6 +1106,9 @@ main(int argc, char **argv)
}
}
 
+   if (pledge("stdio rpath dns inet id proc", NULL) == -1)
+   err(1, "pledge");
+
argv += optind;
argc -= optind;
if ((argc != (ptb->sflag ? 0 : 1)) ||
@@ -1125,6 +1126,9 @@ main(int argc, char **argv)
} else
drop_gid();
 
+   if (pledge("stdio id dns inet", NULL) == -1)
+   err(1, "pledge");
+
if (!ptb->sflag)
host = argv[0];
/*
@@ -1169,6 +1173,9 @@ main(int argc, char **argv)
errx(1, "getaddrinfo: %s", gai_strerror(herr));
}
 
+   if (pledge("stdio id inet", NULL) == -1)
+   err(1, "pledge");
+
if (getrlimit(RLIMIT_NOFILE, ) == -1)
err(1, "getrlimit");
if (rl.rlim_cur < MAX_FD)
@@ -1177,7 +1184,10 @@ main(int argc, char **argv)
err(1, "setrlimit");
if (getrlimit(RLIMIT_NOFILE, ) == -1)
err(1, "getrlimit");
-   
+
+   if (pledge("stdio inet", NULL) == -1)
+   err(1, "pledge");
+
/* Init world */
TAILQ_INIT(_queue);
if ((ptb->dummybuf = malloc(ptb->dummybuf_len)) == NULL)
@@ -1216,7 +1226,10 @@ main(int argc, char **argv)
}
client_init(aitop, nconn, udp_sc, aib);
}
-   
+
+   if (pledge("stdio", NULL) == -1)
+   err(1, "pledge");
+
/* libevent main loop*/
event_dispatch();
 
 

ip6_output bcopy->memcpy

[David Hill]

Hello -

This diff converts some malloc/bcopy to malloc/memcpy.  netinet6
didn't get the bcopy->memcpy overhaul like netinet did.

Index: netinet6/ip6_output.c
===
RCS file: /cvs/src/sys/netinet6/ip6_output.c,v
retrieving revision 1.189
diff -u -p -r1.189 ip6_output.c
--- netinet6/ip6_output.c   23 Sep 2015 08:49:46 -  1.189
+++ netinet6/ip6_output.c   8 Oct 2015 15:09:01 -
@@ -2189,7 +2189,7 @@ do {\
dst->type = malloc(hlen, M_IP6OPT, canwait);\
if (dst->type == NULL && canwait == M_NOWAIT)\
goto bad;\
-   bcopy(src->type, dst->type, hlen);\
+   memcpy(dst->type, src->type, hlen);\
}\
 } while (/*CONSTCOND*/ 0)
 
@@ -2211,7 +2211,7 @@ copypktopts(struct ip6_pktopts *dst, str
M_IP6OPT, canwait);
if (dst->ip6po_nexthop == NULL)
goto bad;
-   bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
+   memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
src->ip6po_nexthop->sa_len);
}
PKTOPT_EXTHDRCPY(ip6po_hbh);
@@ -2847,7 +2847,7 @@ ip6_setpktopt(int optname, u_char *buf, 
opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
if (opt->ip6po_nexthop == NULL)
return (ENOBUFS);
-   bcopy(buf, opt->ip6po_nexthop, *buf);
+   memcpy(opt->ip6po_nexthop, buf, *buf);
break;
 
case IPV6_2292HOPOPTS:
@@ -2882,7 +2882,7 @@ ip6_setpktopt(int optname, u_char *buf, 
opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
if (opt->ip6po_hbh == NULL)
return (ENOBUFS);
-   bcopy(hbh, opt->ip6po_hbh, hbhlen);
+   memcpy(opt->ip6po_hbh, hbh, hbhlen);
 
break;
}
@@ -2945,7 +2945,7 @@ ip6_setpktopt(int optname, u_char *buf, 
*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
if (*newdest == NULL)
return (ENOBUFS);
-   bcopy(dest, *newdest, destlen);
+   memcpy(*newdest, dest, destlen);
 
break;
}
@@ -2986,7 +2986,7 @@ ip6_setpktopt(int optname, u_char *buf, 
opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
if (opt->ip6po_rthdr == NULL)
return (ENOBUFS);
-   bcopy(rth, opt->ip6po_rthdr, rthlen);
+   memcpy(opt->ip6po_rthdr, rth, rthlen);
break;
}
 

Re: kill struct inpcbtable's inpt_lastport

[David Hill]

On Sat, Sep 19, 2015 at 07:57:00PM +0200, Vincent Gross wrote:
> On 09/18/15 23:39, David Hill wrote:
> > On Fri, Sep 18, 2015 at 11:05:55PM +0200, Vincent Gross wrote:
> >> On 09/18/15 15:18, David Hill wrote:
> >>> Is this 'if (count)' statement needed?  We know first > last, so count
> >>> will always be positive.  lastport will always be set.
> >>
> >>> if last == first, then the if statement will be false and lastport will
> >>> be uninitialized, I believe.
> >>>
> >>
> >> Both remarks are true, but I think it is better to keep a more extensive
> >> refactoring in a separate diff, refactoring that shall get rid of this
> >> yucky code duplication.
> >>
> > 
> > Well, this code changes the current behavior.  I'd at least change
> > lastport to be initialized to 0 to keep the behavior the same.  It was
> > previously set to 0 with M_ZERO.
> > 
> 
> Fixed. Ok ?

LGTM.

> 
> Index: sys/netinet/in_pcb.c
> ===
> RCS file: /cvs/src/sys/netinet/in_pcb.c,v
> retrieving revision 1.179
> diff -u -p -r1.179 in_pcb.c
> --- sys/netinet/in_pcb.c  11 Sep 2015 15:29:47 -  1.179
> +++ sys/netinet/in_pcb.c  19 Sep 2015 17:52:42 -
> @@ -199,7 +199,6 @@ in_pcbinit(struct inpcbtable *table, int
>   >inpt_lhash);
>   if (table->inpt_lhashtbl == NULL)
>   panic("in_pcbinit: hashinit failed for lport");
> - table->inpt_lastport = 0;
>   table->inpt_count = 0;
>   arc4random_buf(>inpt_key, sizeof(table->inpt_key));
>  }
> @@ -281,8 +280,8 @@ in_pcbbind(struct inpcb *inp, struct mbu
>  {
>   struct socket *so = inp->inp_socket;
>   struct inpcbtable *table = inp->inp_table;
> - u_int16_t *lastport = >inp_table->inpt_lastport;
>   struct sockaddr_in *sin;
> + u_int16_t lastport = 0;
>   u_int16_t lport = 0;
>   int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
>   int error;
> @@ -391,16 +390,16 @@ in_pcbbind(struct inpcb *inp, struct mbu
>*/
>   count = first - last;
>   if (count)
> - *lastport = first - arc4random_uniform(count);
> + lastport = first - arc4random_uniform(count);
>  
>   do {
>   if (count-- < 0)/* completely used? */
>   return (EADDRNOTAVAIL);
> - --*lastport;
> - if (*lastport > first || *lastport < last)
> - *lastport = first;
> - lport = htons(*lastport);
> - } while (in_baddynamic(*lastport, 
> so->so_proto->pr_protocol) ||
> + --lastport;
> + if (lastport > first || lastport < last)
> + lastport = first;
> + lport = htons(lastport);
> + } while (in_baddynamic(lastport, 
> so->so_proto->pr_protocol) ||
>   in_pcblookup(table, _addr, 0,
>   >inp_laddr, lport, wild, inp->inp_rtableid));
>   } else {
> @@ -409,16 +408,16 @@ in_pcbbind(struct inpcb *inp, struct mbu
>*/
>   count = last - first;
>   if (count)
> - *lastport = first + arc4random_uniform(count);
> + lastport = first + arc4random_uniform(count);
>  
>   do {
>   if (count-- < 0)/* completely used? */
>   return (EADDRNOTAVAIL);
> - ++*lastport;
> - if (*lastport < first || *lastport > last)
> - *lastport = first;
> - lport = htons(*lastport);
> - } while (in_baddynamic(*lastport, 
> so->so_proto->pr_protocol) ||
> + ++lastport;
> + if (lastport < first || lastport > last)
> + lastport = first;
> + lport = htons(lastport);
> + } while (in_baddynamic(lastport, 
> so->so_proto->pr_protocol) ||
>   in_pcblookup(table, _addr, 0,
>

Re: kill struct inpcbtable's inpt_lastport

[David Hill]

On Fri, Sep 18, 2015 at 11:05:55PM +0200, Vincent Gross wrote:
> On 09/18/15 15:18, David Hill wrote:
> > Is this 'if (count)' statement needed?  We know first > last, so count
> > will always be positive.  lastport will always be set.
> 
> > if last == first, then the if statement will be false and lastport will
> > be uninitialized, I believe.
> > 
> 
> Both remarks are true, but I think it is better to keep a more extensive
> refactoring in a separate diff, refactoring that shall get rid of this
> yucky code duplication.
> 

Well, this code changes the current behavior.  I'd at least change
lastport to be initialized to 0 to keep the behavior the same.  It was
previously set to 0 with M_ZERO.
 

> --
> Vincent Gross
>