On Saturday 30 December 2006 15:11, Karol Kwiatkowski wrote: > Ivan Voras wrote: > > Karol Kwiatkowski wrote: > >> This works: FreeBSD 6.2-PRERELEASE #0: Thu Dec 14 11:34:36 CET 2006 > >> This doesn't: FreeBSD 6.2-PRERELEASE #0: Sat Dec 30 11:27:11 CET > >> 2006 > >> > >> I'm not sure it that's all that matters, I can supply more > >> information if needed. > > > > Yes, you'll need to send at least a kernel stack backtrace. See here: > > http://www.freebsd.org/doc/en_US.ISO8859-1/books/developers-handbook/ > >kerneldebug-online-ddb.html > > > > Get it to start the debugger on panic and post the backtrace (bt). > > Thanks for the link, I hope I've done it right. Here it is: > > > Fatal trap 12: page fault while in kernel mode > fault virtual address = 0x74 > fault code = supervisor read, page not present > instruction pointer = 0x20:0xc055d18d > stack pointer = 0x28:0xe989dbbc > frame pointer = 0x28:0xe989dbc0 > code segment = base 0x0, limit 0xfffff, type 0x1b > = DPL 0, pres 1, def32 1, gran 1 > processor eflags = resume, IOPL = 0 > current process = 637 (ntpd) > [thread pid 637 tid 100059 ] > Stopped at turnstile_setowner+0xd: movl 0x74(%ecx),%eax > db> bt > Tracing pid 637 tid 100059 td 0xc4de2780 > turnstile_setowner(c4de9240,0,c4dc9240,c07201e0,c5102090,...) at > turnstile_setowner+0xd > turnstile_wait(c5101090,0,c5101000,c07234e0,e989dc24,...) at > turnstile_wait+0xca > _mtx_lock_sleep(c5102090,c4de2780,0,0,0,...) at _mtx_lock_sleep+0xb4 > in_pcballoc(c4e892c8,c07234e0,10000,c06c8de5,38,...) at > in_pcballoc+0xbe tcp_attach(c4e892c8,c4e8933c,c06c8de5,0,c4e892c8,...) > at tcp_attach+0x58 tcp_usr_attach(c4e892c8,0,c4de2780,0,0,...) at > tcp_usr_attach+0x63 socreate(2,e989dcb8,1,0,c4ab1c90,...) at > socreate+0x167 > socket(c4de2780,e989dd04,c,c4de2780,80d3000,...) at socket+0xb3 > syscall(3b,3b,3b,0,7,...) at syscall+0x380 > Xint0x80_syscall() at Xint0x80_syscall+0x1f > --- syscall (97, FreeBSD ELF32, socket), eip = 0x282939d3, esp = > 0xbfbfeb1c, ebp = 0xbfbfebf8 --- > db >
Something like the attached should fix it. Seems tcp was forgotten in a recent change to INP locking. -- /"\ Best regards, | [EMAIL PROTECTED] \ / Max Laier | ICQ #67774661 X http://pf4freebsd.love2party.net/ | [EMAIL PROTECTED] / \ ASCII Ribbon Campaign | Against HTML Mail and News
Index: tcp_subr.c
===================================================================
RCS file: /usr/store/mlaier/fcvs/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.228.2.12
diff -u -r1.228.2.12 tcp_subr.c
--- tcp_subr.c 1 Oct 2006 05:33:50 -0000 1.228.2.12
+++ tcp_subr.c 30 Dec 2006 15:06:45 -0000
@@ -300,6 +300,15 @@
uma_zone_set_max(tcptw_zone, tcptw_auto_size());
}
+static int
+tcp_inpcb_init(void *mem, int size, int flags)
+{
+ struct inpcb *inp = mem;
+
+ INP_LOCK_INIT(inp, "inp", "tcpinp");
+ return (0);
+}
+
void
tcp_init()
{
@@ -328,7 +337,7 @@
tcbinfo.porthashbase = hashinit(hashsize, M_PCB,
&tcbinfo.porthashmask);
tcbinfo.ipi_zone = uma_zcreate("inpcb", sizeof(struct inpcb),
- NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
+ NULL, NULL, tcp_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
uma_zone_set_max(tcbinfo.ipi_zone, maxsockets);
#ifdef INET6
#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
pgpTdZroeH8uw.pgp
Description: PGP signature
