> --- a/fs/exec.c
> +++ b/fs/exec.c
> @@ -1413,6 +1413,9 @@ int do_execve(const char * filename,
> free_bprm(bprm);
> if (displaced)
> put_files_struct(displaced);
> + /* Just for testing. */
> + if (!retval)
> + retval = 1;
> return retval;
>
> out:
>
> With a patch above (on x86 CentOS 5.5 with 2.6.36-rc6 kernel),
> a few programs failed to work properly.
>
> udevd-event[$PID]: run_program: '$PROGRAM' abnormal exit
>
I found below messages in dmesg when above error message is printed.
[ 16.082683] usb_id[2097]: segfault at dfa68ce8 ip 0806ffb7 sp bf892b30 error 5 in usb_id[8048000+cd000]
[ 16.732016] scsi_id[2133]: segfault at dfa37e78 ip 08071157 sp bf85d640 error 5 in scsi_id[8048000+cf000]
[ 16.754163] scsi_id[2138]: segfault at e002ca08 ip 08071157 sp bfe521d0 error 4 in scsi_id[8048000+cf000]
[ 16.910293] scsi_id[2162]: segfault at e008def8 ip 08071157 sp bfeb36c0 error 4 in scsi_id[8048000+cf000]
[ 16.971028] scsi_id[2167]: segfault at dff9f908 ip 08071157 sp bfdc50d0 error 5 in scsi_id[8048000+cf000]
[ 17.435741] vol_id[2177]: segfault at e017d258 ip 08072bb7 sp bff9c0a0 error 4 in vol_id[8048000+d0000]
[ 17.497793] vol_id[2179]: segfault at dfc82ab8 ip 08072bb7 sp bfaa1900 error 5 in vol_id[8048000+d0000]
[ 17.676152] edd_id[2197]: segfault at dff0de58 ip 0806f097 sp bfd3b920 error 5 in edd_id[8048000+ca000]
[ 17.678064] edd_id[2196]: segfault at dff55a38 ip 0806f097 sp bfd83500 error 5 in edd_id[8048000+ca000]
[ 18.317277] vol_id[2210]: segfault at e00c8a48 ip 08072bb7 sp bfee7890 error 4 in vol_id[8048000+d0000]
And I confirmed that applying below patch in addition to above patch solves
these segfault failures.
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -315,7 +315,7 @@ long sys_execve(const char __user *name,
error = do_execve(filename, argv, envp, regs);
#ifdef CONFIG_X86_32
- if (error == 0) {
+ if (error >= 0) {
/* Make sure we don't return using sysenter.. */
set_thread_flag(TIF_IRET);
}
This result indicates that sys_execve() expected that do_execve() returns 0
if do_execve() successfully replaced the current process's image.
Now, the question is "when is do_execve() > 0 allowed"?
If do_execve() > 0 is allowed when the current process's image was successfully
replaced, we need to either "modify do_execve() callers not to assume 0 on
success" or "modify do_execve() return 0 on success".
Regards.
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
