Re: [PATCH dwarves] btf_encoder: always initialize func_state to 0

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 







On Friday, January 10th, 2025 at 5:48 AM, Arnaldo Carvalho de Melo <acme@xxxxxxxxxx> wrote:

> 
> 
> On Fri, Jan 10, 2025 at 10:39:50AM +0000, Alan Maguire wrote:
> 
> > On 10/01/2025 02:31, Ihor Solodrai wrote:
> > 
> > > BPF CI caught a segfault on aarch64 and s390x [1] after recent merges
> > > into the master branch.
> > > 
> > > The segfault happened at free(func_state->annots) in
> > > btf_encoder__delete_saved_funcs().
> > > 
> > > func_state->annots arrived there uninitialized because after patch [2]
> > > in some cases func_state may be allocated with a realloc, but was not
> > > zeroed out.
> > > 
> > > Fix this bug by always memset-ing a func_state to zero in
> > > btf_encoder__alloc_func_state().
> > > 
> > > [1] https://github.com/kernel-patches/bpf/actions/runs/12700574327
> > > [2] https://lore.kernel.org/dwarves/20250109185950.653110-11-ihor.solodrai@xxxxx/
> > 
> > Thanks for the quick fix! Reproduced this on an aarch64 system:
> > 
> > BTF [M] kernel/resource_kunit.ko
> > /bin/sh: line 1: 630875 Segmentation fault (core dumped)
> > LLVM_OBJCOPY="objcopy" pahole -J -j
> > --btf_features=encode_force,var,float,enum64,decl_tag,type_tag,optimized_func,consistent_func,decl_tag_kfuncs
> > --lang_exclude=rust --btf_base ./vmlinux kernel/kcsan/kcsan_test.ko
> > make[2]: *** [scripts/Makefile.modfinal:57: kernel/kcsan/kcsan_test.ko]
> > Error 139
> > make[2]: *** Deleting file 'kernel/kcsan/kcsan_test.ko'
> > make[2]: *** Waiting for unfinished jobs....
> > /bin/sh: line 1: 630907 Segmentation fault (core dumped)
> > LLVM_OBJCOPY="objcopy" pahole -J -j
> > --btf_features=encode_force,var,float,enum64,decl_tag,type_tag,optimized_func,consistent_func,decl_tag_kfuncs
> > --lang_exclude=rust --btf_base ./vmlinux kernel/torture.ko
> > make[2]: *** [scripts/Makefile.modfinal:56: kernel/torture.ko] Error 139
> > make[2]: *** Deleting file 'kernel/torture.ko'
> > 
> > ...and verified that with the fix all works well.
> > 
> > Nit: missing Signed-off-by
> 
> 
> I'm adding it, ok?

Yes, thanks!

Signed-off-by: Ihor Solodrai <ihor.solodrai@xxxxx>

> 
> - Arnaldo
> 
> > Tested-by: Alan Maguire alan.maguire@xxxxxxxxxx
> 
> 
> Thanks!
> 
> > > ---
> > > btf_encoder.c | 7 +++++--
> > > 1 file changed, 5 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/btf_encoder.c b/btf_encoder.c
> > > index 78efd70..511c1ea 100644
> > > --- a/btf_encoder.c
> > > +++ b/btf_encoder.c
> > > @@ -1083,7 +1083,7 @@ static bool funcs__match(struct btf_encoder_func_state *s1,
> > > 
> > > static struct btf_encoder_func_state *btf_encoder__alloc_func_state(struct btf_encoder *encoder)
> > > {
> > > - struct btf_encoder_func_state *tmp;
> > > + struct btf_encoder_func_state *state, *tmp;
> > > 
> > > if (encoder->func_states.cnt >= encoder->func_states.cap) {
> > > 
> > > @@ -1100,7 +1100,10 @@ static struct btf_encoder_func_state *btf_encoder__alloc_func_state(struct btf_e
> > > encoder->func_states.array = tmp;
> > > }
> > > 
> > > - return &encoder->func_states.array[encoder->func_states.cnt++];
> > > + state = &encoder->func_states.array[encoder->func_states.cnt++];
> > > + memset(state, 0, sizeof(*state));
> > > +
> > > + return state;
> > > }
> > > 
> > > static int32_t btf_encoder__save_func(struct btf_encoder *encoder, struct function *fn, struct elf_function *func)





[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux