Date: Mon, 30 Mar 2015 12:05:41 -0500 From: Alan Cox <alc@rice.edu> To: Don Lewis <truckman@FreeBSD.org> Cc: src-committers@FreeBSD.org, alc@FreeBSD.org, svn-src-all@FreeBSD.org, bdrewery@FreeBSD.org, portmgr@FreeBSD.org, svn-src-head@FreeBSD.org, clusteradm@FreeBSD.org Subject: Re: svn commit: r280327 - in head/sys: kern vm Message-ID: <55198265.7050809@rice.edu> In-Reply-To: <201503300625.t2U6PE3c093114@gw.catspoiler.org> References: <201503300625.t2U6PE3c093114@gw.catspoiler.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 03/30/2015 01:25, Don Lewis wrote: > On 28 Mar, Alan Cox wrote: >> On 03/28/2015 12:29, Bryan Drewery wrote: >>> On 3/27/2015 9:41 PM, Don Lewis wrote: >>>> On 21 Mar, Alan Cox wrote: >>>>> Author: alc >>>>> Date: Sat Mar 21 17:56:55 2015 >>>>> New Revision: 280327 >>>>> URL: https://svnweb.freebsd.org/changeset/base/280327 >>>>> >>>>> Log: >>>>> Introduce vm_object_color() and use it in mmap(2) to set the colo= r of >>>>> named objects to zero before the virtual address is selected. Pr= eviously, >>>>> the color setting was delayed until after the virtual address was= >>>>> selected. In rtld, this delay effectively prevented the mapping = of a >>>>> shared library's code section using superpages. Now, for example= , we see >>>>> the first 1 MB of libc's code on armv6 mapped by a superpage afte= r we've >>>>> gotten through the initial cold misses that bring the first 1 MB = of code >>>>> into memory. (With the page clustering that we perform on read f= aults, >>>>> this happens quickly.) >>>>> =20 >>>>> Differential Revision: https://reviews.freebsd.org/D2013 >>>>> Reviewed by: jhb, kib >>>>> Tested by: Svatopluk Kraus (armv6) >>>>> MFC after: 6 weeks >>>>> >>>>> Modified: >>>>> head/sys/kern/kern_exec.c >>>>> head/sys/vm/vm_fault.c >>>>> head/sys/vm/vm_mmap.c >>>>> head/sys/vm/vm_object.h >>>>> head/sys/vm/vnode_pager.c >>>> This change appears to have partially broken package building. >>>> >>>> I recently set up a package building machine running 11.0-CURRENT >>>> r280642 and found that it was unable to build openjdk7 packages insi= de >>>> FreeBSD 8.4 and 9.3 poudriere jails, for both i386 and amd64. >>>> >>>> [snip] >>>> Compiling ../generated/adfiles/ad_x86_64_gen.cpp >>>> rm -f ad_x86_64_gen.o >>>> c++ -D_ALLBSD_SOURCE -D_GNU_SOURCE -DAMD64 -DPRODUCT -I. -I/wrkdirs/= usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/prims -I/wrkdir= s/usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm -I/wrkdirs/us= r/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/precompiled -I/wr= kdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/cpu/x86/vm -I/wrkd= irs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os_cpu/bsd_x86/vm -I= /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os/bsd/vm -I/wr= kdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os/posix/vm -I../g= enerated -DHOTSPOT_RELEASE_VERSION=3D"\"24.76-b04\"" -DHOTSPOT_BUILD_TARG= ET=3D"\"product\"" -DHOTSPOT_BUILD_USER=3D"\"root\"" -DHOTSPOT_LIB_ARCH=3D= \"amd64\" -DHOTSPOT_VM_DISTRO=3D"\"OpenJDK\"" -O2 -pipe -fstack-protector= -fno-strict-aliasing -DTARGET_OS_FAMILY_bsd -DTARGET_ARCH_x86 -DTARGET_A= RCH_MODEL_x86_64 -DTARGET_OS_ARCH_bsd_x86 -DTARGET_OS_ARCH_MODEL_bsd_x86_= 64 -DTARGET_COMPILER_gcc -DCOMPILER2 -DCOMPILER1 ! > -fno! >>>> -rtti -fno-exceptions -pthread -fcheck-new -m64 -pipe -DTARGET_OS_F= AMILY_bsd -DTARGET_ARCH_x86 -DTARGET_ARCH_MODEL_x86_64 -DTARGET_OS_ARCH_b= sd_x86 -DTARGET_OS_ARCH_MODEL_bsd_x86_64 -DTARGET_COMPILER_gcc -DCOMPILER= 2 -DCOMPILER1 -fPIC -fno-rtti -fno-exceptions -pthread -fcheck-new -m64 -= pipe -O3 -fno-strict-aliasing -DVM_LITTLE_ENDIAN -D_LP64=3D1 -fno-omit-fr= ame-pointer -DINCLUDE_TRACE=3D1 -Wpointer-arith -Wconversion -Wsign-comp= are -c -fpch-deps -MMD -MP -MF ../generated/dependencies/ad_x86_64_fo= rmat.o.d -o ad_x86_64_format.o ../generated/adfiles/ad_x86_64_format.cpp = >>>> c++ -D_ALLBSD_SOURCE -D_GNU_SOURCE -DAMD64 -DPRODUCT -I. -I/wrkdirs/= usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/prims -I/wrkdir= s/usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm -I/wrkdirs/us= r/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/precompiled -I/wr= kdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/cpu/x86/vm -I/wrkd= irs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os_cpu/bsd_x86/vm -I= /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os/bsd/vm -I/wr= kdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os/posix/vm -I../g= enerated -DHOTSPOT_RELEASE_VERSION=3D"\"24.76-b04\"" -DHOTSPOT_BUILD_TARG= ET=3D"\"product\"" -DHOTSPOT_BUILD_USER=3D"\"root\"" -DHOTSPOT_LIB_ARCH=3D= \"amd64\" -DHOTSPOT_VM_DISTRO=3D"\"OpenJDK\"" -O2 -pipe -fstack-protector= -fno-strict-aliasing -DTARGET_OS_FAMILY_bsd -DTARGET_ARCH_x86 -DTARGET_A= RCH_MODEL_x86_64 -DTARGET_OS_ARCH_bsd_x86 -DTARGET_OS_ARCH_MODEL_bsd_x86_= 64 -DTARGET_COMPILER_gcc -DCOMPILER2 -DCOMPILER1 ! > -fno! >>>> -rtti -fno-exceptions -pthread -fcheck-new -m64 -pipe -DTARGET_OS_F= AMILY_bsd -DTARGET_ARCH_x86 -DTARGET_ARCH_MODEL_x86_64 -DTARGET_OS_ARCH_b= sd_x86 -DTARGET_OS_ARCH_MODEL_bsd_x86_64 -DTARGET_COMPILER_gcc -DCOMPILER= 2 -DCOMPILER1 -fPIC -fno-rtti -fno-exceptions -pthread -fcheck-new -m64 -= pipe -O3 -fno-strict-aliasing -DVM_LITTLE_ENDIAN -D_LP64=3D1 -fno-omit-fr= ame-pointer -DINCLUDE_TRACE=3D1 -Wpointer-arith -Wconversion -Wsign-comp= are -c -fpch-deps -MMD -MP -MF ../generated/dependencies/ad_x86_64_ge= n.o.d -o ad_x86_64_gen.o ../generated/adfiles/ad_x86_64_gen.cpp=20 >>>> Compiling ../generated/adfiles/ad_x86_64_misc.cpp >>>> rm -f ad_x86_64_misc.o >>>> c++ -D_ALLBSD_SOURCE -D_GNU_SOURCE -DAMD64 -DPRODUCT -I. -I/wrkdirs/= usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/prims -I/wrkdir= s/usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm -I/wrkdirs/us= r/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/precompiled -I/wr= kdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/cpu/x86/vm -I/wrkd= irs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os_cpu/bsd_x86/vm -I= /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os/bsd/vm -I/wr= kdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/os/posix/vm -I../g= enerated -DHOTSPOT_RELEASE_VERSION=3D"\"24.76-b04\"" -DHOTSPOT_BUILD_TARG= ET=3D"\"product\"" -DHOTSPOT_BUILD_USER=3D"\"root\"" -DHOTSPOT_LIB_ARCH=3D= \"amd64\" -DHOTSPOT_VM_DISTRO=3D"\"OpenJDK\"" -O2 -pipe -fstack-protector= -fno-strict-aliasing -DTARGET_OS_FAMILY_bsd -DTARGET_ARCH_x86 -DTARGET_A= RCH_MODEL_x86_64 -DTARGET_OS_ARCH_bsd_x86 -DTARGET_OS_ARCH_MODEL_bsd_x86_= 64 -DTARGET_COMPILER_gcc -DCOMPILER2 -DCOMPILER1 ! > -fno! >>>> -rtti -fno-exceptions -pthread -fcheck-new -m64 -pipe -DTARGET_OS_F= AMILY_bsd -DTARGET_ARCH_x86 -DTARGET_ARCH_MODEL_x86_64 -DTARGET_OS_ARCH_b= sd_x86 -DTARGET_OS_ARCH_MODEL_bsd_x86_64 -DTARGET_COMPILER_gcc -DCOMPILER= 2 -DCOMPILER1 -fPIC -fno-rtti -fno-exceptions -pthread -fcheck-new -m64 -= pipe -O3 -fno-strict-aliasing -DVM_LITTLE_ENDIAN -D_LP64=3D1 -fno-omit-fr= ame-pointer -DINCLUDE_TRACE=3D1 -Wpointer-arith -Wconversion -Wsign-comp= are -c -fpch-deps -MMD -MP -MF ../generated/dependencies/ad_x86_64_mi= sc.o.d -o ad_x86_64_misc.o ../generated/adfiles/ad_x86_64_misc.cpp=20 >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/c= ompiler/abstractCompiler.cpp:1: fatal error: had to relocate PCH >>>> compilation terminated. >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/src/share/vm/u= tilities/accessFlags.cpp:1: fatal error: had to relocate PCH >>>> compilation terminated. >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'abstractCompiler.o' failed >>>> gmake[6]: *** [abstractCompiler.o] Error 1 >>>> gmake[6]: *** Waiting for unfinished jobs.... >>>> ../generated/adfiles/ad_x86_64.cpp:1: fatal error: had to relocate P= CH >>>> compilation terminated. >>> Are you using ccache? >>> >>>> ../generated/adfiles/ad_x86_64_clone.cpp:1: fatal error: had to relo= cate PCH >>>> compilation terminated. >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'accessFlags.o' failed >>>> gmake[6]: *** [accessFlags.o] Error 1 >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'ad_x86_64_clone.o' failed >>>> gmake[6]: *** [ad_x86_64_clone.o] Error 1 >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'ad_x86_64.o' failed >>>> gmake[6]: *** [ad_x86_64.o] Error 1 >>>> ../generated/adfiles/ad_x86_64_expand.cpp:1: fatal error: had to rel= ocate PCH >>>> compilation terminated. >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'ad_x86_64_expand.o' failed >>>> gmake[6]: *** [ad_x86_64_expand.o] Error 1 >>>> ../generated/adfiles/ad_x86_64_gen.cpp:1: fatal error: had to reloca= te PCH >>>> compilation terminated. >>>> ../generated/adfiles/ad_x86_64_misc.cpp:1: fatal error: had to reloc= ate PCH >>>> compilation terminated. >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'ad_x86_64_gen.o' failed >>>> gmake[6]: *** [ad_x86_64_gen.o] Error 1 >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'ad_x86_64_misc.o' failed >>>> gmake[6]: *** [ad_x86_64_misc.o] Error 1 >>>> ../generated/adfiles/ad_x86_64_format.cpp:1: fatal error: had to rel= ocate PCH >>>> compilation terminated. >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/rules.make:149: recipe for target 'ad_x86_64_format.o' failed >>>> gmake[6]: *** [ad_x86_64_format.o] Error 1 >>>> gmake[6]: Leaving directory '/wrkdirs/usr/ports/java/openjdk7/work/o= penjdk/build/bsd-amd64/hotspot/outputdir/bsd_amd64_compiler2/product' >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/makef= iles/top.make:128: recipe for target 'the_vm' failed >>>> gmake[5]: *** [the_vm] Error 2 >>>> gmake[5]: Leaving directory '/wrkdirs/usr/ports/java/openjdk7/work/o= penjdk/build/bsd-amd64/hotspot/outputdir/bsd_amd64_compiler2/product' >>>> /wrkdirs/usr/ports/java/openjdk7/work/openjdk/hotspot/make/bsd/Makef= ile:292: recipe for target 'product' failed >>>> gmake[4]: *** [product] Error 2 >>>> gmake[4]: Leaving directory '/wrkdirs/usr/ports/java/openjdk7/work/o= penjdk/build/bsd-amd64/hotspot/outputdir' >>>> Makefile:191: recipe for target 'generic_build2' failed >>>> gmake[3]: *** [generic_build2] Error 2 >>>> gmake[3]: Leaving directory '/wrkdirs/usr/ports/java/openjdk7/work/o= penjdk/hotspot/make' >>>> Makefile:151: recipe for target 'product' failed >>>> gmake[2]: *** [product] Error 2 >>>> gmake[2]: Leaving directory '/wrkdirs/usr/ports/java/openjdk7/work/o= penjdk/hotspot/make' >>>> make/hotspot-rules.gmk:111: recipe for target 'hotspot-build' failed= >>>> gmake[1]: *** [hotspot-build] Error 2 >>>> gmake[1]: Leaving directory '/wrkdirs/usr/ports/java/openjdk7/work/o= penjdk' >>>> Makefile:251: recipe for target 'build_product_image' failed >>>> gmake: *** [build_product_image] Error 2 >>>> =3D=3D=3D> Compilation failed unexpectedly. >>>> >>>> I was not seeing this problem on my older package builder running >>>> 10.1-STABLE. Since this problem has not shown up on the FreeBSD pac= kage >>>> building cluster, I got suspicious that the change was quite recent.= >>>> >>>> This old gcc bug report: >>>> <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D14940>; led me to sus= pect >>>> mmap(). >>>> >>>> The old gcc source file /usr/src/contrib/gcc/ggc-common.c does a cou= ple >>>> of mmap() calls. Tne first is in mmap_gt_pch_get_address() where a= >>>> NULL first argument is used. The address that gets returned is stas= hed >>>> away and the region is unmapped. Then a later call in >>>> mmap_gt_pch_use_address() passes this saved address to mmap() as a >>>> hint. It expects the mapped region to get mapped to the same base >>>> address. If this does not happen, the above error is the result. >>> I don't know what I'm talking about but that doesn't sound like a ver= y >>> safe assumption for the gcc code to make. >> >> Your intuition is correct. It is not. And, in fact, under Solaris an= d >> Linux, gcc does not make this assumption. >> >> I suspect that the solution used by gcc under Solaris would work for >> us. However, that would entail modifying gcc in older branches. > After digging in to this today, I'm not sure that the Solaris solution > would work. Actually, based on what you've written below, the solution in config/host-solaris.c should work. That solution uses mincore(2) to verify that the desired addresses are not in use and reissues the mmap(2) call with MAP_FIXED if they are not. MAP_FIXED overrides the automatic realignment that moves the mapping up to 0x8115F4000. This should work on any version of FreeBSD. In FreeBSD 10+, we have a new flag MAP_EXCL that eliminates the need for the mincore(2) calls and the second call to mmap(2). The first call can safely specify MAP_EXCL | MAP_FIXED. > On amd64, with an amd64 jail, and kernel rev r280326, I observe the > following: > One gcc process calls mmap() with addr=3D0 and len=3D0x657a000, and > the value 0x811400000 is returned. Subsequent gcc processes > call mmap() with addr=3D0x811400000 and len=3D0x657a000, gettin= g > 0x811400000 in return. > > With kernel rev r280327 I see: > One gcc process calls mmap() with addr=3D0 and len=3D0x657a000, and > the value 0x811400000 is returned. Subsequent gcc processes > call mmap() with addr=3D0x811400000 and len=3D0x657a000, getting > 0x8115f4000 in return. What I later noticed is that the subsequent > calls are passing offset=3D0x1f4000. Not so coincidentally, > 0x811400000 + 0x1f4000 =3D 0x8115F4000. > > My first attempt at a fix subtracted offset from address, but the mmap(= ) > return value changed to 0x8113f4000 instead of the 0x811400000 I was > expecting. It looked to me like the code must be doing superpage > alignment on the start of the file and then adding the offset to get th= e > start of the mapped region. > > This somewhat hacking patch disables this alignment if a non-zero > address is passed as a hint, and allows the code to make the start of > the mapped region match the hint. With this patch, I've been able to > build openjdk7 in a FreeBSD 9.3 amd64 jail. > > Index: vm_mmap.c > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > --- vm_mmap.c (revision 280327) > +++ vm_mmap.c (working copy) > @@ -325,6 +325,8 @@ > * There should really be a pmap call to determine a reasonable > * location. > */ > + if (addr !=3D 0 && (flags & MAP_ANON) =3D=3D 0) > + flags |=3D MAP_RESERVED0100; > PROC_LOCK(td->td_proc); > if (addr =3D=3D 0 || > (addr >=3D round_page((vm_offset_t)vms->vm_taddr) && > @@ -1665,6 +1667,8 @@ > else if ((flags & MAP_ALIGNMENT_MASK) !=3D 0) > findspace =3D VMFS_ALIGNED_SPACE(flags >> > MAP_ALIGNMENT_SHIFT); > + else if ((flags & MAP_RESERVED0100) !=3D 0) > + findspace =3D VMFS_ANY_SPACE; > else > findspace =3D VMFS_OPTIMAL_SPACE; > rv =3D vm_map_find(map, object, foff, addr, size, > > =20 >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?55198265.7050809>