ptrace still broken- PTRACE_CONT trap flag, and debug state

["In Cognito" <defend.the.world@gmail.com>]

Hello,

main(){
__asm__("pushf\r\n"
              "popf\r\n"
             );
}

Single stepping through the "pushf" instruction and then attempting to
PTRACE_CONT will generate SIGTRAPs at each instruction. Although the
trap flag appears to not be set according to GETREG values, the pushf
will push a value with the TF bit on, which is then restored with
popf...

Here's an example of single stepping on a 2.6.8 and the trap flag
(uppercase letter means enabled):

8048351: push %ecx.
 eax=bffff7cc ebx=40146adc ecx=bffff740 edx=00000001
 esi=bffff7c4 edi=bffff750 ebp=bffff728 esp=bffff724
 flags= [c] [P] [a] [z] [S] [T] [I] [df] [id]

8048352: pushf .
 eax=bffff7cc ebx=40146adc ecx=bffff740 edx=00000001
 esi=bffff7c4 edi=bffff750 ebp=bffff728 esp=bffff720
 flags= [c] [P] [a] [z] [S] [T] [I] [df] [id]

8048353: popf .
 eax=bffff7cc ebx=40146adc ecx=bffff740 edx=00000001
 esi=bffff7c4 edi=bffff750 ebp=bffff728 esp=bffff724
 flags= [c] [P] [a] [z] [S] [T] [I] [df] [id]

8048354: pop %ecx.
 eax=bffff7cc ebx=40146adc ecx=bffff740 edx=00000001
 esi=bffff7c4 edi=bffff750 ebp=bffff728 esp=bffff728
 flags= [c] [P] [a] [z] [S] [T] [I] [df] [id]



and a 2.6.20



8048351: push %ecx.
 eax=bfcff654 ebx=b7f52ff4 ecx=bfcff5d0 edx=00000001
 esi=b7f80ce0 edi=00000000 ebp=bfcff5b8 esp=bfcff5b4
 flags= [c] [P] [a] [z] [S] [t] [I] [df] [id] 286

8048352: pushf .
 eax=bfcff654 ebx=b7f52ff4 ecx=bfcff5d0 edx=00000001
 esi=b7f80ce0 edi=00000000 ebp=bfcff5b8 esp=bfcff5b0
 flags= [c] [P] [a] [z] [S] [t] [I] [df] [id] 286

8048353: popf .
 eax=bfcff654 ebx=b7f52ff4 ecx=bfcff5d0 edx=00000001
 esi=b7f80ce0 edi=00000000 ebp=bfcff5b8 esp=bfcff5b4
 flags= [c] [P] [a] [z] [S] [T] [I] [df] [id] 386

8048354: pop %ecx.
 eax=bfcff654 ebx=b7f52ff4 ecx=bfcff5d0 edx=00000001
 esi=b7f80ce0 edi=00000000 ebp=bfcff5b8 esp=bfcff5b8
 flags= [c] [P] [a] [z] [S] [T] [I] [df] [id] 386



I'm not sure where along the line this was broken but it looks like
there were more than a few changes...
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/