> What's the best way to fix it ? fix kauth_cred_geteuid(), or audit the
> kauth_cred_geteuid() calls and handle it there ?
I guess it would be better if NOCRED/FSCRED is handled in
kauth_cred_geteuid() (and other kauth_cred_... routines).
In case of kauth_cred_geteuid(), it will be natual if 0 i
Manuel Bouyer writes:
> Stopped in pid 9987.1 (postdrop) at netbsd:kauth_cred_geteuid+0xd:
> movl 4
> 4(%rbx),%eax
> kauth_cred_geteuid() at netbsd:kauth_cred_geteuid+0xd
> ffs_alloc() at netbsd:ffs_alloc+0x1aa
> ffs_balloc() at netbsd:ffs_balloc+0x1525
> wrsnapblk() at netbsd:wrsnapblk+0x4f
FSC
> I don't think it is from a quick reading. The only reason it works,
> is because most of the time it rounds up.
It looks like the only place where the member is used is here:
tmpfs_vnops.c: error = uiomove(node->tn_spec.tn_lnk.tn_link,
tmpfs_vnops.c- MIN(node->tn_size, uio->uio_resi
Martin Husemann writes:
> Are there any ET_DYN that are not shared libs? Maybe we should remove
> the is_dyn exec support completely.
It looks like some linux binary is ET_DYN rather than ET_EXEC and
exec_elfNN_makecmds is used by comapt code also.
$ readelf -h /usr/bin/man
ELF Header:
Magic:
Maxime Villard writes:
>> - How often is_dyn is true?
>
> Most of the time, I think.
Hm, that's different from what I've expected.
> Whether it is true or not doesn't
> really matter, does it ?
It matters if possibility of "is_dyn is true case" and
"elf_check_header(eh, ET_EXEC) == 0 is true c
Maxime Villard writes:
> * Put is_dyn before. It's just a small optimization:
>elf_check_header(eh, ET_EXEC) is always called before checking
>is_dyn, so if we invert the two things we avoid calling
>elf_check_header() twice if is_dyn is ok.
- How often is_dyn is true?
- Isn't it b