Re: buildkernel broken for META_MODE
> > [Creating objdir /usr/obj/usr/obj/root/git/freebsd/sys/GENERIC...] > Wrong^ > > Note we have 'cd /usr/obj/' and 'MAKEOBJDIRPREFIX=/usr/obj' in > there, so we get a nested /usr/obj/.CURDIR problem of /usr/obj/usr/obj. The following would probably help that case: Index: auto.obj.mk === RCS file: /volume/cvs/sjg/mk/auto.obj.mk,v retrieving revision 1.13 diff -u -p -r1.13 auto.obj.mk --- auto.obj.mk 24 Mar 2017 20:53:22 - 1.13 +++ auto.obj.mk 18 Apr 2017 22:25:08 - @@ -41,6 +41,10 @@ MKOBJDIRS= auto # Use __objdir here so it is easier to tweak without impacting # the logic. .if !empty(MAKEOBJDIRPREFIX) +.if ${.CURDIR:M${MAKEOBJDIRPREFIX}/*} != "" +# we are already in obj tree! +__objdir?= ${.CURDIR} +.endif __objdir?= ${MAKEOBJDIRPREFIX}${.CURDIR} .endif __objdir?= ${MAKEOBJDIR:Uobj} ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
Re: buildkernel broken for META_MODE
Bryan Drewerywrote: > > What is the issue above? diff? > > I don't know what the issue is with buildkernel specifically, I never > looked into it. Buildworld had other issues like rescue/rescue not being > AUTO_OBJ safe. That's fixed. I forget the details of buildworld as well. > One of the simpler changes is to remove the 'make obj' treewalk and > prefer .OBJDIR/OBJTOP over other ${CANONICALOBJDIR} ${OBJTREE} type > hacks. My patches are quite mangled with the idea of moving to > /usr/obj/CURDIR/TARGET.TARGET_ARCH/RELDIR and I need to split it all up > still. Ok > > > > > FWIW the latest version of auto.obj.mk deals with __objdir=obj which is > > a corner case compared to any use of MAKEOJDIRPREFIX etc. > > > > It does not fix the issues. Not this lot no. Its for folk wanting to use the traditional ${.CURDIR}/obj/ which used to throw an error. I haven't used ${.CURDIR}/obj/ in almost 20 years ;-) > > SIZE="size" INSTALL="sh /root/git/freebsd/tools/install.sh" > > PATH=/usr/obj/root/git/freebsd/tmp/legacy/usr/sbin:/usr/obj/root/git/freebsd/tmp/legacy/usr/bin:/usr/obj/root/git/freebsd/tmp/legacy/bin:/usr/obj/root/git/freebsd/tmp/usr/sbin:/usr/obj/root/git/freebsd/tmp/usr/bin:/sbin:/bin:/usr/sbin:/usr/bin > > make -j 1 > > 5 -D NO_FILEMON -d M -J 15,16 -m /root/git/freebsd/share/mk KERNEL=kernel > > obj > > [Creating objdir /usr/obj/usr/obj/root/git/freebsd/sys/GENERIC...] > Wrong^ > > Note we have 'cd /usr/obj/' and 'MAKEOBJDIRPREFIX=/usr/obj' in > there, so we get a nested /usr/obj/.CURDIR problem of /usr/obj/usr/obj. Yes, ugly. ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
Re: buildkernel broken for META_MODE
On 4/18/2017 2:33 PM, Simon J. Gerraty wrote: > Bryan Drewerywrote: >> Aha /usr/obj/usr/obj. >> >> That was in Renato's report as well. >> >> The bug is WITH_AUTO_OBJ. I just confirmed that. A bunch of errors occur >> when doing the first build and the opt_*.h files are not generated in >> the "proper" place by config(8). >> >> WITH_AUTO_OBJ is not really workable yet in buildworld/buildkernel. I >> have patches that I will commit soon to add support for them (meaning to >> work as it all does not but without a 'make obj' tree walk). > > What is the issue above? diff? I don't know what the issue is with buildkernel specifically, I never looked into it. Buildworld had other issues like rescue/rescue not being AUTO_OBJ safe. That's fixed. I forget the details of buildworld as well. One of the simpler changes is to remove the 'make obj' treewalk and prefer .OBJDIR/OBJTOP over other ${CANONICALOBJDIR} ${OBJTREE} type hacks. My patches are quite mangled with the idea of moving to /usr/obj/CURDIR/TARGET.TARGET_ARCH/RELDIR and I need to split it all up still. > > FWIW the latest version of auto.obj.mk deals with __objdir=obj which is > a corner case compared to any use of MAKEOJDIRPREFIX etc. > It does not fix the issues. > -- stage 2.2: rebuilding the object tree > -- > cd /usr/obj/root/git/freebsd/sys/GENERIC; COMPILER_VERSION=4 > COMPILER_FEATURES=c++11 COMPILER_TYPE=clang > COMPILER_FREEBSD_VERSION=126 MAKEOBJDIRPREFIX=/usr/obj > MACHINE_ARCH=amd64 MACHINE=amd64 CPUTYPE= BUILD_TOOLS_META=.NOMETA > GROFF_BIN_PATH=/usr/obj/root/git/freebsd/tmp/legacy/usr/bin GROFF_FONT_PATH > =/usr/obj/root/git/freebsd/tmp/legacy/usr/share/groff_font > GROFF_TMAC_PATH=/usr/obj/root/git/freebsd/tmp/legacy/usr/share/tmac > CC="/usr/local/bin/ccache cc -target x86_64-unknown-freebsd12.0 > --sysroot=/usr/obj/root/git/freebsd/tmp > -B/usr/obj/root/git/freebsd/tmp/usr/bin" CXX="/usr/local/bin/ccache c++ > -target x86_ > 64-unknown-freebsd12.0 --sysroot=/usr/obj/root/git/freebsd/tmp > -B/usr/obj/root/git/freebsd/tmp/usr/bin" CPP="cpp -target > x86_64-unknown-freebsd12.0 --sysroot=/usr/obj/root/git/freebsd/tmp > -B/usr/obj/root/git/freebsd/tmp/usr/bin" AS="as" AR="ar" LD="ld" > LLVM_LINK="" NM=nm OBJCOPY="objcopy" RANLIB=ranlib STRINGS= > SIZE="size" INSTALL="sh /root/git/freebsd/tools/install.sh" > PATH=/usr/obj/root/git/freebsd/tmp/legacy/usr/sbin:/usr/obj/root/git/freebsd/tmp/legacy/usr/bin:/usr/obj/root/git/freebsd/tmp/legacy/bin:/usr/obj/root/git/freebsd/tmp/usr/sbin:/usr/obj/root/git/freebsd/tmp/usr/bin:/sbin:/bin:/usr/sbin:/usr/bin > make -j 1 > 5 -D NO_FILEMON -d M -J 15,16 -m /root/git/freebsd/share/mk KERNEL=kernel obj > [Creating objdir /usr/obj/usr/obj/root/git/freebsd/sys/GENERIC...] Wrong^ Note we have 'cd /usr/obj/' and 'MAKEOBJDIRPREFIX=/usr/obj' in there, so we get a nested /usr/obj/.CURDIR problem of /usr/obj/usr/obj. > grep: opt_md.h: No such file or directory > grep: opt_ddb.h: No such file or directory > grep: opt_kdtrace.h: No such file or directory > grep: opt_hwpmc_hooks.h: No such file or directory Wrong^ I will look into this as part of my larger patch, but it's likely due to the first bug. -- Regards, Bryan Drewery signature.asc Description: OpenPGP digital signature
Re: https://reviews.freebsd.org/D10232
On Tue, 2017-04-18 at 23:15 +0300, Toomas Soome wrote: > Hi! > > The network code rework still needs good people to test out the work, > especially the non-x86 platforms:) the url is still: https://reviews > .freebsd.org/D10232 > > Once it checks out, we can go for next phase and get some boost for > network loading. > > thanks, > toomas I use netbooting in loader(8) extensively for armv6, and in fact that's what I'm knee-deep in for $work today too, so I'll give D10232 a look tonight or tomorrow at the latest. -- Ian ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
Re: buildkernel broken for META_MODE
Bryan Drewerywrote: > Aha /usr/obj/usr/obj. > > That was in Renato's report as well. > > The bug is WITH_AUTO_OBJ. I just confirmed that. A bunch of errors occur > when doing the first build and the opt_*.h files are not generated in > the "proper" place by config(8). > > WITH_AUTO_OBJ is not really workable yet in buildworld/buildkernel. I > have patches that I will commit soon to add support for them (meaning to > work as it all does not but without a 'make obj' tree walk). What is the issue above? diff? FWIW the latest version of auto.obj.mk deals with __objdir=obj which is a corner case compared to any use of MAKEOJDIRPREFIX etc. ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
https://reviews.freebsd.org/D10232
Hi! The network code rework still needs good people to test out the work, especially the non-x86 platforms:) the url is still: https://reviews.freebsd.org/D10232 Once it checks out, we can go for next phase and get some boost for network loading. thanks, toomas ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
Re: buildkernel broken for META_MODE
On 4/10/2017 11:14 AM, Bryan Drewery wrote: > On 4/7/2017 12:13 PM, Ngie Cooper (yaneurabeya) wrote: >> Hi, >> I ran into this error when trying to run a meta mode build (for the >> first time). It might be related to the recent assym* ordering changes. >> Thanks! >> -Ngie >> >> $ cat /etc/src-env.conf >> WITH_AUTO_OBJ= yes Oh I missed this part^ >> WITH_META_MODE= yes >> UPDATE_DEPENDFILE= yes By the way UPDATE_DEPENDFILE is not really a user variable and setting it may break things. >> $ pwd >> /usr/src >> $ svnversion >> 316603M >> $ svn st | grep -v \? >> M usr.bin/grep/tests/Makefile >> $ env SRCCONF=/dev/null NO_FILEMON=1 script ~/bk.ts make buildkernel -j3 >> Script started on Fri Apr 7 11:52:38 2017 >> Command: time make buildkernel -j3 >> --- buildkernel --- >> make[1]: "/usr/src/Makefile.inc1" line 146: SYSTEM_COMPILER: Determined that >> CC=cc matches the source tree. Not bootstrapping a cross-compiler. >> --- buildkernel --- >> -- > Kernel build for GENERIC started on Fri Apr 7 11:52:39 PDT 2017 >> -- >> ===> GENERIC >> mkdir -p /usr/obj/usr/src/sys >> -- > stage 1: configuring the kernel >> -- >> ... >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/acpi_quirks.h Aha /usr/obj/usr/obj. That was in Renato's report as well. The bug is WITH_AUTO_OBJ. I just confirmed that. A bunch of errors occur when doing the first build and the opt_*.h files are not generated in the "proper" place by config(8). WITH_AUTO_OBJ is not really workable yet in buildworld/buildkernel. I have patches that I will commit soon to add support for them (meaning to work as it all does not but without a 'make obj' tree walk). >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/feeder_eq_gen.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/feeder_rate_gen.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/snd_fxdiv_gen.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/miidevs.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/pccarddevs.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/teken_state.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/usbdevs.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/vnode_if.h >> Building /usr/obj/usr/obj/usr/src/sys/GENERIC/ia32_genassym.o >> --- ia32_genassym.o --- >> :1:10: fatal error: 'opt_global.h' file not found >> #include "opt_global.h" >> ^~ >> 1 error generated. >> *** [ia32_genassym.o] Error code 1 >> >> make[2]: stopped in /usr/obj/usr/src/sys/GENERIC >> .ERROR_TARGET='ia32_genassym.o' >> .ERROR_META_FILE='/usr/obj/usr/obj/usr/src/sys/GENERIC/ia32_genassym.o.meta' >> > > It was previously reported here too > https://lists.freebsd.org/pipermail/freebsd-current/2016-June/061946.html. > It's a race. I had forgotten about it but will look at it again. > -- Regards, Bryan Drewery signature.asc Description: OpenPGP digital signature
Re: Sendmail eats CPU on r317039 [after -r316874 it may be -r316951 and -r316973 are not enough to fix everything]
Ronald Klop ronald-lists at klop.ws wrote on Tue Apr 18 09:59:50 UTC 2017: > there is a thread ono this list about a problem in syslogd which made > syslog-clients (like sendmail busy-looping on logging. > That might be related to this. (it is fixed in the source already, so > upgrading again might help) > See the thread with subject like 'Re: r316958: booting a server takes >10 > minutes!' > > Regards, > > Ronald. Yes. But Tom V.'s report is for -r317039, which is after the reported fixes as far as I can tell. Something besides syslogd might also cause problems? In my nearly-default -r317015 ardm64 context [as a VirtualBox guest] I've not seen the problem, where I did before. (The only reason sendmail runs in my context is for the messages FreeBSD sends to it own local accounts. I do not otherwise use mail in this context.) Tom V.'s report vs. others finding lack of a problem suggests that the coverage of the fixes is incomplete somehow but useful. I happen to not be doing whatever causes the problem to appear. I've no clue what might be different or unusual in Tom V.'s context. There is also the possibility that Tom V.'s report is a fully independent issue. But such does not seem all that likely on the initial information. > On 2017-Apr-17, at 7:57 AM, Mark Millard wrote: > >> Just an FYI of a more recent report of runaway sendmail on a >> more recent system version ( -r317039 ): >> >> Begin forwarded message: >> >>> From: Tom Vijlbrief >>> Subject: Sendmail eats CPU on r317039 >>> Date: April 17, 2017 at 3:39:37 AM PDT >>> To: "freebsd-current at freebsd.org" , >>> freebsd-arm >>> >>> On a recent kernel sendmail is constantly consuming CPU. >>> >>> truss -p PID >>> >>> shows: >>> >>> sendto(3,"<22>Apr 17 10:30:33 sendmail[362"...,163,0x0,NULL,0) ERR#55 'No >>> buffer space available' >>> nanosleep({ 0.01000 }) = 0 (0x0) >>> sendto(3,"<22>Apr 17 10:30:33 sendmail[362"...,163,0x0,NULL,0) ERR#55 'No >>> buffer space available' >>> nanosleep({ 0.01000 }) >>> ... >>> >>> This is on an arm64 system >> >> Analysis of Tom V.'s context for this may be required. === Mark Millard markmi at dsl-only.net ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
Re: Fwd: Sendmail eats CPU on r317039 [after -r316874 it may be -r316951 and -r316973 are not enough to fix everything]
there is a thread ono this list about a problem in syslogd which made syslog-clients (like sendmail busy-looping on logging. That might be related to this. (it is fixed in the source already, so upgrading again might help) See the thread with subject like 'Re: r316958: booting a server takes >10 minutes!' Regards, Ronald. On Mon, 17 Apr 2017 16:57:39 +0200, Mark Millardwrote: Just an FYI of a more recent report of runaway sendmail on a more recent system version ( -r317039 ): Begin forwarded message: From: Tom Vijlbrief Subject: Sendmail eats CPU on r317039 Date: April 17, 2017 at 3:39:37 AM PDT To: "freebsd-current at freebsd.org" , freebsd-arm On a recent kernel sendmail is constantly consuming CPU. truss -p PID shows: sendto(3,"<22>Apr 17 10:30:33 sendmail[362"...,163,0x0,NULL,0) ERR#55 'No buffer space available' nanosleep({ 0.01000 }) = 0 (0x0) sendto(3,"<22>Apr 17 10:30:33 sendmail[362"...,163,0x0,NULL,0) ERR#55 'No buffer space available' nanosleep({ 0.01000 }) ... This is on an arm64 system Analysis of Tom V.'s context for this may be required. === Mark Millard markmi at dsl-only.net ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org" ___ freebsd-current@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"