Skip to contentSkip to conversation

2.6.15-rc1-mm2 breaks strace

DISCUSSION Bernhard Rosenkraenzer 3 participants 3 messages Nov 18, 2005 ptrace support

lore
Share & tools
lore
#0 lore raw

2.6.15-rc1-mm2 works nicely here, aside from the artsd stuff someone else already reported, and an new issue with strace (last known working:
2.6.14-mm2)

[bero@localhost bero]$ strace ls
execve("/bin/ls", ["ls"], [/* 33 vars */]) = 0 Segmentation fault
[ls output without any traces beyond execve-ing ls displayed here]

Also interesting:
[bero@localhost bero]$ strace strace ls execve("/usr/bin/strace", ["strace", "ls"], [/* 20 vars */]) = 0 Segmentation fault
execve("/bin/ls", ["ls"], [/* 20 vars */]) = 0
[ls output without any traces displayed here]

#1 lore raw
Quoted 3 lines · Bernhard Rosenkraenzer

2.6.15-rc1-mm2 works nicely here, aside from the artsd stuff someone else already reported, and an new issue with strace (last known working:
2.6.14-mm2)

I think, that the problem emerged in rc1-mm1, there are some changes in ptrace.c (2.6.14-mm2 works fine for me).

Quoted 6 lines

[bero@localhost bero]$ strace ls
execve("/bin/ls", ["ls"], [/* 33 vars */]) = 0 Segmentation fault
[ls output without any traces beyond execve-ing ls displayed here]

Quoted 6 lines

Also interesting:
[bero@localhost bero]$ strace strace ls execve("/usr/bin/strace", ["strace", "ls"], [/* 20 vars */]) = 0 Segmentation fault
execve("/bin/ls", ["ls"], [/* 20 vars */]) = 0
[ls output without any traces displayed here]

And I have this in logs:

Unable to handle kernel NULL pointer dereference at virtual address 00000010 printing eip:
c0125a25
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP 
last sysfs file: /devices/platform/i2c-9191/9191-0290/temp3_input
Modules linked in: usblp it87 hwmon_vid i2c_isa sunrpc ipv6 video ohci1394 ieee1394 emu10k1_gp gameport ide_cd cdrom
CPU:    0
EIP:    0060:[<c0125a25>]    Not tainted VLI
EFLAGS: 00210206   (2.6.15-rc1-mm2) 

EIP is at ptrace_check_attach+0x19/0xb8
eax: 00ffffff ebx: 00000000 ecx: c5ea0b3c edx: c5ea0b38
esi: 00000000 edi: 00000000 ebp: c6106f90 esp: c6106f80
ds: 007b es: 007b ss: 0068
Process strace (pid: 23486, threadinfo=c6106000 task=c17ba550)

Stack: c01262dc 00000000 00000000 00000018 c6106fb4 c0126344 00000000 00000000 
       00000000 c6106fbc 00000018 00000000 5000cff4 c6106000 c0102e15 00000018 
       00005bbf 00000001 00000000 5000cff4 bf8ced48 ffffffda 0000007b c010007b 
Call Trace:
 [<c0103c82>] show_stack+0x9a/0xd0
 [<c0103e42>] show_registers+0x16a/0x1fa
 [<c010407c>] die+0x123/0x1ae
 [<c03c3b3c>] do_page_fault+0x33c/0x66e
 [<c010393b>] error_code+0x4f/0x54
 [<c0126344>] sys_ptrace+0x5a/0xcf
 [<c0102e15>] syscall_call+0x7/0xb
Code: cc ba 42 c0 c7 43 68 c8 ba 42 c0 89 50 04 89 02 eb a4 55 89 e5 57 56 53 83 ec 04 8b 75 08 8b 7d 0c b8 00 2d 4c c0 e8 16 d5 29 00 <8b> 56 10 f6 c2 01 75 25 bb fd ff ff ff b8 00 2d 4c c0 e8 65 d7 
 <3>BUG: strace[23486] exited with nonzero preempt_count 1!

regards,
--
Jiri Slaby www.fi.muni.cz/~xslaby
\_.-^-._ [email protected] _.-^-._/
B67499670407CE62ACC8 22A032CC55C339D47A7E

#2 lore raw
In reply to Bernhard Rosenkraenzer
Quoted 9 lines · Bernhard Rosenkraenzer

2.6.15-rc1-mm2 works nicely here, aside from the artsd stuff someone else already reported, and an new issue with strace (last known working:
2.6.14-mm2)

[bero@localhost bero]$ strace ls
execve("/bin/ls", ["ls"], [/* 33 vars */]) = 0 Segmentation fault
[ls output without any traces beyond execve-ing ls displayed here]

Quoted 6 lines

Also interesting:
[bero@localhost bero]$ strace strace ls execve("/usr/bin/strace", ["strace", "ls"], [/* 20 vars */]) = 0 Segmentation fault
execve("/bin/ls", ["ls"], [/* 20 vars */]) = 0
[ls output without any traces displayed here]

Try Christoph's patch.

Thanks,
Badari

Looks like 2.6.15-rc1-mm1 has total crap in ptrace_get_task_struct (and it looks like my fault because I sent out a wrong patch).

The patch below should fix it:

kernel/ptrace.c +1 -1
Index: linux-2.6/kernel/ptrace.c
===================================================================
@@ -459,7 +459,7 @@
 	read_unlock(&tasklist_lock);
 	if (!child)
 		return ERR_PTR(-ESRCH);
-	return 0;
+	return child;
 }
 
 #ifndef __ARCH_SYS_PTRACE