Message ID | 20200317211649.o4fzaxrzy6qxvz4f@google.com |
---|---|
State | Changes Requested |
Delegated to: | BPF Maintainers |
Headers | show |
Series | [bpf-next,v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF | expand |
On 03/17, Fangrui Song wrote: > Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump. > The existing 'file format' and 'architecture' parsing logic is brittle > and does not work with llvm-objcopy/llvm-objdump. Thanks, it all makes sense and looks much better/portable (too much dependence on binutils :-). I've left a bunch of questions/nits below. > > .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and > rename .BTF to BTF so that C code can reference the section via linker > synthesized __start_BTF and __stop_BTF. This fixes a small problem that > previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary` > synthesized symbols _binary__btf_vmlinux_bin_start and > _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with > more common __start_BTF and __stop_BTF. > > Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns > "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?" > > We use a dd command to change the e_type field in the ELF header from > ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o. Accepting > ET_EXEC as an input file is an extremely rare GNU ld feature that lld > does not intend to support, because this is error-prone. Please keep small changelog here, for example: v5: * rebased on top of bpfnext Btw, I tried to pull and test it and failed: $ curl -LO https://lore.kernel.org/bpf/20200317211649.o4fzaxrzy6qxvz4f@google.com/raw % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 9627 100 9627 0 0 47191 0 --:--:-- --:--:-- --:--:-- 47191 $ git am raw warning: Patch sent with format=flowed; space at the end of lines might be lost. Applying: bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF error: corrupt patch at line 20 Patch failed at 0001 bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF hint: Use 'git am --show-current-patch' to see the failed patch When you have resolved this problem, run "git am --continue". If you prefer to skip this patch, run "git am --skip" instead. To restore the original branch and stop patching, run "git am --abort". $ git describe v5.6-rc5-1621-g230021539e8c Are you sure it's on top of the bpf-next? Or am I doing something wrong? > Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux") > Cc: Stanislav Fomichev <sdf@google.com> > Cc: Nick Desaulniers <ndesaulniers@google.com> > Reported-by: Nathan Chancellor <natechancellor@gmail.com> > Link: https://github.com/ClangBuiltLinux/linux/issues/871 > Signed-off-by: Fangrui Song <maskray@google.com> > --- > kernel/bpf/btf.c | 9 ++++----- > kernel/bpf/sysfs_btf.c | 11 +++++------ > scripts/link-vmlinux.sh | 17 ++++++----------- > 3 files changed, 15 insertions(+), 22 deletions(-) > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c > index 50080add2ab9..6f397c4da05e 100644 > --- a/kernel/bpf/btf.c > +++ b/kernel/bpf/btf.c > @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size, > return ERR_PTR(err); > } > -extern char __weak _binary__btf_vmlinux_bin_start[]; > -extern char __weak _binary__btf_vmlinux_bin_end[]; > +extern char __weak __start_BTF[]; > +extern char __weak __stop_BTF[]; > extern struct btf *btf_vmlinux; > #define BPF_MAP_TYPE(_id, _ops) > @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void) > } > env->btf = btf; > - btf->data = _binary__btf_vmlinux_bin_start; > - btf->data_size = _binary__btf_vmlinux_bin_end - > - _binary__btf_vmlinux_bin_start; > + btf->data = __start_BTF; > + btf->data_size = __stop_BTF - __start_BTF; > err = btf_parse_hdr(env); > if (err) > diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c > index 7ae5dddd1fe6..3b495773de5a 100644 > --- a/kernel/bpf/sysfs_btf.c > +++ b/kernel/bpf/sysfs_btf.c > @@ -9,15 +9,15 @@ > #include <linux/sysfs.h> > /* See scripts/link-vmlinux.sh, gen_btf() func for details */ > -extern char __weak _binary__btf_vmlinux_bin_start[]; > -extern char __weak _binary__btf_vmlinux_bin_end[]; > +extern char __weak __start_BTF[]; > +extern char __weak __stop_BTF[]; > static ssize_t > btf_vmlinux_read(struct file *file, struct kobject *kobj, > struct bin_attribute *bin_attr, > char *buf, loff_t off, size_t len) > { > - memcpy(buf, _binary__btf_vmlinux_bin_start + off, len); > + memcpy(buf, __start_BTF + off, len); > return len; > } > @@ -30,15 +30,14 @@ static struct kobject *btf_kobj; > static int __init btf_vmlinux_init(void) > { > - if (!_binary__btf_vmlinux_bin_start) > + if (!__start_BTF) > return 0; > btf_kobj = kobject_create_and_add("btf", kernel_kobj); > if (!btf_kobj) > return -ENOMEM; > - bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end - > - _binary__btf_vmlinux_bin_start; > + bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF; > return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux); > } > diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh > index ac569e197bfa..ae2048625f1e 100755 > --- a/scripts/link-vmlinux.sh > +++ b/scripts/link-vmlinux.sh > @@ -133,17 +133,12 @@ gen_btf() > info "BTF" ${2} > LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1} > - # dump .BTF section into raw binary file to link with final vmlinux > - bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \ > - cut -d, -f1 | cut -d' ' -f2) > - bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \ > - awk '{print $4}') > - bin_file=.btf.vmlinux.bin We still have the following in gen_btf that you need to remove: local bin_arch local bin_format local bin_file > - ${OBJCOPY} --change-section-address .BTF=0 \ > - --set-section-flags .BTF=alloc -O binary \ > - --only-section=.BTF ${1} $bin_file > - ${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \ > - --rename-section .data=.BTF $bin_file ${2} > + # Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference > + # it via linker synthesized __start_BTF and __stop_BTF. Change e_type > + # to ET_REL so that it can be used to link final vmlinux. > + ${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \ > + --rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \ > + printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none We have 'set -e' so && can be omitted. Maybe it will be a bit more clear: # Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference # it via linker synthesized __start_BTF and __stop_BTF. ${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \ --rename-section .BTF=BTF ${1} ${2} 2>/dev/null # Change e_type to ET_REL (0x01) so that it can be used to link final # vmlinux. printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none ^^^ maybe also point out that this is required for llvm's ldd. I know you point it out in the commit message, but still nice to have as a comment here. > } > # Create ${2} .o file with all symbols from the ${1} object file > -- > 2.25.1.481.gfbce0eb801-goog
On Tue, Mar 17, 2020 at 02:16:49PM -0700, Fangrui Song wrote: > Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump. > The existing 'file format' and 'architecture' parsing logic is brittle > and does not work with llvm-objcopy/llvm-objdump. > > .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and > rename .BTF to BTF so that C code can reference the section via linker > synthesized __start_BTF and __stop_BTF. This fixes a small problem that > previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary` > synthesized symbols _binary__btf_vmlinux_bin_start and > _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with > more common __start_BTF and __stop_BTF. I'm glad to see the name change benefit here. Just reducing the number of execs in this path is only worth this change, IMO. Going from 2 objdump and 2 objcopy calls to a single objcopy is very nice. :) > > Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns > "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?" > > We use a dd command to change the e_type field in the ELF header from > ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o. Accepting > ET_EXEC as an input file is an extremely rare GNU ld feature that lld > does not intend to support, because this is error-prone. Nice ELF trick. :) > > Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux") > Cc: Stanislav Fomichev <sdf@google.com> > Cc: Nick Desaulniers <ndesaulniers@google.com> > Reported-by: Nathan Chancellor <natechancellor@gmail.com> > Link: https://github.com/ClangBuiltLinux/linux/issues/871 > Signed-off-by: Fangrui Song <maskray@google.com> Reviewed-by: Kees Cook <keescook@chromium.org> -Kees > --- > kernel/bpf/btf.c | 9 ++++----- > kernel/bpf/sysfs_btf.c | 11 +++++------ > scripts/link-vmlinux.sh | 17 ++++++----------- > 3 files changed, 15 insertions(+), 22 deletions(-) > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c > index 50080add2ab9..6f397c4da05e 100644 > --- a/kernel/bpf/btf.c > +++ b/kernel/bpf/btf.c > @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size, > return ERR_PTR(err); > } > -extern char __weak _binary__btf_vmlinux_bin_start[]; > -extern char __weak _binary__btf_vmlinux_bin_end[]; > +extern char __weak __start_BTF[]; > +extern char __weak __stop_BTF[]; > extern struct btf *btf_vmlinux; > #define BPF_MAP_TYPE(_id, _ops) > @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void) > } > env->btf = btf; > - btf->data = _binary__btf_vmlinux_bin_start; > - btf->data_size = _binary__btf_vmlinux_bin_end - > - _binary__btf_vmlinux_bin_start; > + btf->data = __start_BTF; > + btf->data_size = __stop_BTF - __start_BTF; > err = btf_parse_hdr(env); > if (err) > diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c > index 7ae5dddd1fe6..3b495773de5a 100644 > --- a/kernel/bpf/sysfs_btf.c > +++ b/kernel/bpf/sysfs_btf.c > @@ -9,15 +9,15 @@ > #include <linux/sysfs.h> > /* See scripts/link-vmlinux.sh, gen_btf() func for details */ > -extern char __weak _binary__btf_vmlinux_bin_start[]; > -extern char __weak _binary__btf_vmlinux_bin_end[]; > +extern char __weak __start_BTF[]; > +extern char __weak __stop_BTF[]; > static ssize_t > btf_vmlinux_read(struct file *file, struct kobject *kobj, > struct bin_attribute *bin_attr, > char *buf, loff_t off, size_t len) > { > - memcpy(buf, _binary__btf_vmlinux_bin_start + off, len); > + memcpy(buf, __start_BTF + off, len); > return len; > } > @@ -30,15 +30,14 @@ static struct kobject *btf_kobj; > static int __init btf_vmlinux_init(void) > { > - if (!_binary__btf_vmlinux_bin_start) > + if (!__start_BTF) > return 0; > btf_kobj = kobject_create_and_add("btf", kernel_kobj); > if (!btf_kobj) > return -ENOMEM; > - bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end - > - _binary__btf_vmlinux_bin_start; > + bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF; > return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux); > } > diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh > index ac569e197bfa..ae2048625f1e 100755 > --- a/scripts/link-vmlinux.sh > +++ b/scripts/link-vmlinux.sh > @@ -133,17 +133,12 @@ gen_btf() > info "BTF" ${2} > LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1} > - # dump .BTF section into raw binary file to link with final vmlinux > - bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \ > - cut -d, -f1 | cut -d' ' -f2) > - bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \ > - awk '{print $4}') > - bin_file=.btf.vmlinux.bin > - ${OBJCOPY} --change-section-address .BTF=0 \ > - --set-section-flags .BTF=alloc -O binary \ > - --only-section=.BTF ${1} $bin_file > - ${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \ > - --rename-section .data=.BTF $bin_file ${2} > + # Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference > + # it via linker synthesized __start_BTF and __stop_BTF. Change e_type > + # to ET_REL so that it can be used to link final vmlinux. > + ${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \ > + --rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \ > + printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none > } > # Create ${2} .o file with all symbols from the ${1} object file > -- > 2.25.1.481.gfbce0eb801-goog
On 2020-03-17, Stanislav Fomichev wrote: >On 03/17, Fangrui Song wrote: >> Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump. >> The existing 'file format' and 'architecture' parsing logic is brittle >> and does not work with llvm-objcopy/llvm-objdump. >Thanks, it all makes sense and looks much better/portable (too much >dependence on binutils :-). >I've left a bunch of questions/nits below. > >> >> .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and >> rename .BTF to BTF so that C code can reference the section via linker >> synthesized __start_BTF and __stop_BTF. This fixes a small problem that >> previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary` >> synthesized symbols _binary__btf_vmlinux_bin_start and >> _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with >> more common __start_BTF and __stop_BTF. >> >> Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns >> "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?" >> >> We use a dd command to change the e_type field in the ELF header from >> ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o. Accepting >> ET_EXEC as an input file is an extremely rare GNU ld feature that lld >> does not intend to support, because this is error-prone. >Please keep small changelog here, for example: > >v5: >* rebased on top of bpfnext Thanks for the tip. Add them at the bottom? >Btw, I tried to pull and test it and failed: >$ curl -LO https://lore.kernel.org/bpf/20200317211649.o4fzaxrzy6qxvz4f@google.com/raw > % Total % Received % Xferd Average Speed Time Time Time Current > Dload Upload Total Spent Left Speed >100 9627 100 9627 0 0 47191 0 --:--:-- --:--:-- --:--:-- 47191 >$ git am raw >warning: Patch sent with format=flowed; space at the end of lines might >be lost. >Applying: bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF >error: corrupt patch at line 20 >Patch failed at 0001 bpf: Support llvm-objcopy and llvm-objdump for >vmlinux BTF >hint: Use 'git am --show-current-patch' to see the failed patch >When you have resolved this problem, run "git am --continue". >If you prefer to skip this patch, run "git am --skip" instead. >To restore the original branch and stop patching, run "git am --abort". >$ git describe >v5.6-rc5-1621-g230021539e8c > >Are you sure it's on top of the bpf-next? Or am I doing something wrong? It can be my fault when messing with various patches.. Will be more cautious. >> Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux") >> Cc: Stanislav Fomichev <sdf@google.com> >> Cc: Nick Desaulniers <ndesaulniers@google.com> >> Reported-by: Nathan Chancellor <natechancellor@gmail.com> >> Link: https://github.com/ClangBuiltLinux/linux/issues/871 >> Signed-off-by: Fangrui Song <maskray@google.com> >> --- >> kernel/bpf/btf.c | 9 ++++----- >> kernel/bpf/sysfs_btf.c | 11 +++++------ >> scripts/link-vmlinux.sh | 17 ++++++----------- >> 3 files changed, 15 insertions(+), 22 deletions(-) >> >> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c >> index 50080add2ab9..6f397c4da05e 100644 >> --- a/kernel/bpf/btf.c >> +++ b/kernel/bpf/btf.c >> @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size, >> return ERR_PTR(err); >> } >> -extern char __weak _binary__btf_vmlinux_bin_start[]; >> -extern char __weak _binary__btf_vmlinux_bin_end[]; >> +extern char __weak __start_BTF[]; >> +extern char __weak __stop_BTF[]; >> extern struct btf *btf_vmlinux; >> #define BPF_MAP_TYPE(_id, _ops) >> @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void) >> } >> env->btf = btf; >> - btf->data = _binary__btf_vmlinux_bin_start; >> - btf->data_size = _binary__btf_vmlinux_bin_end - >> - _binary__btf_vmlinux_bin_start; >> + btf->data = __start_BTF; >> + btf->data_size = __stop_BTF - __start_BTF; >> err = btf_parse_hdr(env); >> if (err) >> diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c >> index 7ae5dddd1fe6..3b495773de5a 100644 >> --- a/kernel/bpf/sysfs_btf.c >> +++ b/kernel/bpf/sysfs_btf.c >> @@ -9,15 +9,15 @@ >> #include <linux/sysfs.h> >> /* See scripts/link-vmlinux.sh, gen_btf() func for details */ >> -extern char __weak _binary__btf_vmlinux_bin_start[]; >> -extern char __weak _binary__btf_vmlinux_bin_end[]; >> +extern char __weak __start_BTF[]; >> +extern char __weak __stop_BTF[]; >> static ssize_t >> btf_vmlinux_read(struct file *file, struct kobject *kobj, >> struct bin_attribute *bin_attr, >> char *buf, loff_t off, size_t len) >> { >> - memcpy(buf, _binary__btf_vmlinux_bin_start + off, len); >> + memcpy(buf, __start_BTF + off, len); >> return len; >> } >> @@ -30,15 +30,14 @@ static struct kobject *btf_kobj; >> static int __init btf_vmlinux_init(void) >> { >> - if (!_binary__btf_vmlinux_bin_start) >> + if (!__start_BTF) >> return 0; >> btf_kobj = kobject_create_and_add("btf", kernel_kobj); >> if (!btf_kobj) >> return -ENOMEM; >> - bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end - >> - _binary__btf_vmlinux_bin_start; >> + bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF; >> return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux); >> } >> diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh >> index ac569e197bfa..ae2048625f1e 100755 >> --- a/scripts/link-vmlinux.sh >> +++ b/scripts/link-vmlinux.sh >> @@ -133,17 +133,12 @@ gen_btf() >> info "BTF" ${2} >> LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1} >> - # dump .BTF section into raw binary file to link with final vmlinux >> - bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \ >> - cut -d, -f1 | cut -d' ' -f2) >> - bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \ >> - awk '{print $4}') >> - bin_file=.btf.vmlinux.bin >We still have the following in gen_btf that you need to remove: > local bin_arch > local bin_format > local bin_file Thanks. Will delete them. >> - ${OBJCOPY} --change-section-address .BTF=0 \ >> - --set-section-flags .BTF=alloc -O binary \ >> - --only-section=.BTF ${1} $bin_file >> - ${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \ >> - --rename-section .data=.BTF $bin_file ${2} >> + # Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference >> + # it via linker synthesized __start_BTF and __stop_BTF. Change e_type >> + # to ET_REL so that it can be used to link final vmlinux. >> + ${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \ >> + --rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \ >> + printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none > We have 'set -e' so && can be omitted. Maybe it will be a bit > more clear: Agree. > # Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference > # it via linker synthesized __start_BTF and __stop_BTF. > ${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \ > --rename-section .BTF=BTF ${1} ${2} 2>/dev/null > > # Change e_type to ET_REL (0x01) so that it can be used to link final > # vmlinux. > printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none > > ^^^ maybe also point out that this is required for llvm's ldd. > I know you point it out in the commit message, but still nice > to have as a comment here. Will mention lld in v6. >> } >> # Create ${2} .o file with all symbols from the ${1} object file >> -- >> 2.25.1.481.gfbce0eb801-goog
On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote: > > On 2020-03-17, Stanislav Fomichev wrote: > >Please keep small changelog here, for example: > > > >v5: > >* rebased on top of bpfnext > > Thanks for the tip. Add them at the bottom? "Below the fold" see this patch I just sent out: https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u grep "Changes" $ git format-patch -v2 HEAD~ $ vim 0001-...patch <manually add changelog "below the fold">
On Tue, Mar 17, 2020 at 3:08 PM Nick Desaulniers <ndesaulniers@google.com> wrote: > > On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote: > > > > On 2020-03-17, Stanislav Fomichev wrote: > > >Please keep small changelog here, for example: > > > > > >v5: > > >* rebased on top of bpfnext > > > > Thanks for the tip. Add them at the bottom? > > "Below the fold" see this patch I just sent out: > https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u > grep "Changes" > > $ git format-patch -v2 HEAD~ > $ vim 0001-...patch > <manually add changelog "below the fold"> BPF subtree prefers the changelog in the commit body, not the comments (iow, before ---). Add them at the end of you message, see, for example: https://lore.kernel.org/bpf/a428fb88-9b53-27dd-a195-497755944921@iogearbox.net/T/ > -- > Thanks, > ~Nick Desaulniers
On Tue, Mar 17, 2020 at 3:13 PM Stanislav Fomichev <sdf@google.com> wrote: > > On Tue, Mar 17, 2020 at 3:08 PM Nick Desaulniers > <ndesaulniers@google.com> wrote: > > > > On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote: > > > > > > On 2020-03-17, Stanislav Fomichev wrote: > > > >Please keep small changelog here, for example: > > > > > > > >v5: > > > >* rebased on top of bpfnext > > > > > > Thanks for the tip. Add them at the bottom? > > > > "Below the fold" see this patch I just sent out: > > https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u > > grep "Changes" > > > > $ git format-patch -v2 HEAD~ > > $ vim 0001-...patch > > <manually add changelog "below the fold"> > BPF subtree prefers the changelog in the commit body, not the comments > (iow, before ---). > Add them at the end of you message, see, for example: > https://lore.kernel.org/bpf/a428fb88-9b53-27dd-a195-497755944921@iogearbox.net/T/ Sigh, every maintainer is a special snowflake. In our tree, you're only allowed to commit on Thursdays under a blood moon. /s But thanks for the note.
On 3/18/20 12:14 AM, Nick Desaulniers wrote: > On Tue, Mar 17, 2020 at 3:13 PM Stanislav Fomichev <sdf@google.com> wrote: >> On Tue, Mar 17, 2020 at 3:08 PM Nick Desaulniers >> <ndesaulniers@google.com> wrote: >>> On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote: >>>> On 2020-03-17, Stanislav Fomichev wrote: >>>>> Please keep small changelog here, for example: >>>>> >>>>> v5: >>>>> * rebased on top of bpfnext >>>> >>>> Thanks for the tip. Add them at the bottom? >>> >>> "Below the fold" see this patch I just sent out: >>> https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u >>> grep "Changes" >>> >>> $ git format-patch -v2 HEAD~ >>> $ vim 0001-...patch >>> <manually add changelog "below the fold"> >> BPF subtree prefers the changelog in the commit body, not the comments >> (iow, before ---). >> Add them at the end of you message, see, for example: >> https://lore.kernel.org/bpf/a428fb88-9b53-27dd-a195-497755944921@iogearbox.net/T/ > > Sigh, every maintainer is a special snowflake. In our tree, you're > only allowed to commit on Thursdays under a blood moon. /s > > But thanks for the note. For every commit into bpf/bpf-next we automatically put a 'Link:' tag pointing to the lore.kernel.org/bpf long-term archive, so there is always an easy way to retrieve additional information or discussions ([0] as one example). So whether you put it above or below the '---' line I personally don't mind either way. All crucial information that helps a reader later on to understand /why/ we ended up exactly with the current change should be part of the main commit log though. [0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=483d7a30f538e2f8addd32aa9a3d2e94ae55fa65
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c index 50080add2ab9..6f397c4da05e 100644 --- a/kernel/bpf/btf.c +++ b/kernel/bpf/btf.c @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size, return ERR_PTR(err); } -extern char __weak _binary__btf_vmlinux_bin_start[]; -extern char __weak _binary__btf_vmlinux_bin_end[]; +extern char __weak __start_BTF[]; +extern char __weak __stop_BTF[]; extern struct btf *btf_vmlinux; #define BPF_MAP_TYPE(_id, _ops) @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void) } env->btf = btf; - btf->data = _binary__btf_vmlinux_bin_start; - btf->data_size = _binary__btf_vmlinux_bin_end - - _binary__btf_vmlinux_bin_start; + btf->data = __start_BTF; + btf->data_size = __stop_BTF - __start_BTF; err = btf_parse_hdr(env); if (err) diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c index 7ae5dddd1fe6..3b495773de5a 100644 --- a/kernel/bpf/sysfs_btf.c +++ b/kernel/bpf/sysfs_btf.c @@ -9,15 +9,15 @@ #include <linux/sysfs.h> /* See scripts/link-vmlinux.sh, gen_btf() func for details */ -extern char __weak _binary__btf_vmlinux_bin_start[]; -extern char __weak _binary__btf_vmlinux_bin_end[]; +extern char __weak __start_BTF[]; +extern char __weak __stop_BTF[]; static ssize_t btf_vmlinux_read(struct file *file, struct kobject *kobj, struct bin_attribute *bin_attr, char *buf, loff_t off, size_t len) { - memcpy(buf, _binary__btf_vmlinux_bin_start + off, len); + memcpy(buf, __start_BTF + off, len); return len; } @@ -30,15 +30,14 @@ static struct kobject *btf_kobj; static int __init btf_vmlinux_init(void) { - if (!_binary__btf_vmlinux_bin_start) + if (!__start_BTF) return 0; btf_kobj = kobject_create_and_add("btf", kernel_kobj); if (!btf_kobj) return -ENOMEM; - bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end - - _binary__btf_vmlinux_bin_start; + bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF; return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux); } diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh index ac569e197bfa..ae2048625f1e 100755 --- a/scripts/link-vmlinux.sh +++ b/scripts/link-vmlinux.sh @@ -133,17 +133,12 @@ gen_btf() info "BTF" ${2} LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1} - # dump .BTF section into raw binary file to link with final vmlinux - bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \ - cut -d, -f1 | cut -d' ' -f2) - bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \ - awk '{print $4}') - bin_file=.btf.vmlinux.bin - ${OBJCOPY} --change-section-address .BTF=0 \ - --set-section-flags .BTF=alloc -O binary \ - --only-section=.BTF ${1} $bin_file - ${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \ - --rename-section .data=.BTF $bin_file ${2} + # Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference + # it via linker synthesized __start_BTF and __stop_BTF. Change e_type + # to ET_REL so that it can be used to link final vmlinux. + ${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \ + --rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \ + printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none } # Create ${2} .o file with all symbols from the ${1} object file
Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump. The existing 'file format' and 'architecture' parsing logic is brittle and does not work with llvm-objcopy/llvm-objdump. .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and rename .BTF to BTF so that C code can reference the section via linker synthesized __start_BTF and __stop_BTF. This fixes a small problem that previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary` synthesized symbols _binary__btf_vmlinux_bin_start and _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with more common __start_BTF and __stop_BTF. Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?" We use a dd command to change the e_type field in the ELF header from ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o. Accepting ET_EXEC as an input file is an extremely rare GNU ld feature that lld does not intend to support, because this is error-prone. Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux") Cc: Stanislav Fomichev <sdf@google.com> Cc: Nick Desaulniers <ndesaulniers@google.com> Reported-by: Nathan Chancellor <natechancellor@gmail.com> Link: https://github.com/ClangBuiltLinux/linux/issues/871 Signed-off-by: Fangrui Song <maskray@google.com> --- kernel/bpf/btf.c | 9 ++++----- kernel/bpf/sysfs_btf.c | 11 +++++------ scripts/link-vmlinux.sh | 17 ++++++----------- 3 files changed, 15 insertions(+), 22 deletions(-)