Re: [PATCH v2] binfmt_misc: pass info about P flag by AT_FLAGS

[Laurent Vivier]

Le 16/01/2020 Ã 09:17, YunQiang Su a ÃcritÂ:
> Laurent Vivier <laurent@xxxxxxxxx> ä2020å1æ16æåå äå4:07åéï
>>
>> Le 16/01/2020 Ã 03:20, YunQiang Su a Ãcrit :
>>> From: YunQiang Su <ysu@xxxxxxxxxxxx>
>>>
>>> Currently program invoked by binfmt_misc cannot be aware about whether
>>> P flag, aka preserve path is enabled.
>>>
>>> Some applications like qemu need to know since it has 2 use case:
>>>   1. call by hand, like: qemu-mipsel-static test.app OPTION
>>>      so, qemu have to assume that P option is not enabled.
>>>   2. call by binfmt_misc. If qemu cannot know about whether P flag is
>>>      enabled, distribution's have to set qemu without P flag, and
>>>      binfmt_misc call qemu like:
>>>        qemu-mipsel-static /absolute/path/to/test.app OPTION
>>>      even test.app is not called by absoulute path, like
>>>        ./relative/path/to/test.app
>>>
>>> This patch passes this information by the 3rd bits of unused AT_FLAGS.
>>> Then, in qemu, we can get this info by:
>>>    getauxval(AT_FLAGS) & (1<<3)
>>>
>>> v1->v2:
>>>   not enable kdebug
>>>
>>> See: https://bugs.launchpad.net/qemu/+bug/1818483
>>> Signed-off-by: YunQiang Su <ysu@xxxxxxxxxxxx>
>>> ---
>>>  fs/binfmt_elf.c         | 6 +++++-
>>>  fs/binfmt_elf_fdpic.c   | 6 +++++-
>>>  fs/binfmt_misc.c        | 2 ++
>>>  include/linux/binfmts.h | 4 ++++
>>>  4 files changed, 16 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c
>>> index f4713ea76e82..d33ee07d7f57 100644
>>> --- a/fs/binfmt_elf.c
>>> +++ b/fs/binfmt_elf.c
>>> @@ -178,6 +178,7 @@ create_elf_tables(struct linux_binprm *bprm, const struct elfhdr *exec,
>>>       unsigned char k_rand_bytes[16];
>>>       int items;
>>>       elf_addr_t *elf_info;
>>> +     elf_addr_t flags = 0;
>>>       int ei_index;
>>>       const struct cred *cred = current_cred();
>>>       struct vm_area_struct *vma;
>>> @@ -252,7 +253,10 @@ create_elf_tables(struct linux_binprm *bprm, const struct elfhdr *exec,
>>>       NEW_AUX_ENT(AT_PHENT, sizeof(struct elf_phdr));
>>>       NEW_AUX_ENT(AT_PHNUM, exec->e_phnum);
>>>       NEW_AUX_ENT(AT_BASE, interp_load_addr);
>>> -     NEW_AUX_ENT(AT_FLAGS, 0);
>>> +     if (bprm->interp_flags & BINPRM_FLAGS_PRESERVE_ARGV0) {
>>> +             flags |= BINPRM_FLAGS_PRESERVE_ARGV0;
>>> +     }
>>
>> Perhaps we also need a different flag in AT_FLAG than in interp_flag as
>> BINPRM_FLAGS_PRESERVE_ARGV0 is also part of the internal ABI?
> 
> yep. It may be really a problem.
> So, should we define a set of new macros for AT_FLAGS?

Yes, I think.

Thanks,
Laurent