itrace patches

itrace patches

[Dave Nomura]

Are the two patches that I submitted OK to commit?

What is the procedure for getting patches committed.  In other 
OpenSource communities there are a select few with permission do do commits.

-- 
Dave Nomura
LTC Linux Power Toolchain

Re: itrace patches

[Frank Ch. Eigler]

Dave Nomura <dcnltc@us.ibm.com> writes:

> Are the two patches that I submitted OK to commit?
> [...]

Thanks for committing.  I'll move the test suite file to the
appropriate place.  While testing that though, I found a problem, so
I'll initially disable the test.  Can you elaborate upon the level of
testing of this code?  Has it been run under lockdep?


On an RHEL5 x86-64 machine, I got:

BUG: soft lockup - CPU#0 stuck for 10s! [ls_15964:16126]
CPU 0:
Modules linked in: stap_7c70ff0a5aa33e45a459d750475957f4_763(U) xt_tcpudp iptable_nat ip_nat ip_conntrack nfnet
link ip_tables x_tables xfrm_nalgo crypto_api hidp l2cap bluetooth nfs lockd fscache nfs_acl sunrpc dm_multipat
h video sbs backlight i2c_ec button battery asus_acpi acpi_memhotplug ac lp floppy ata_piix libata scsi_mod i2c
_piix4 i2c_core pcspkr ide_cd 8139too parport_pc 8139cp mii parport serio_raw cdrom dm_snapshot dm_zero dm_mirr
or dm_mod ext3 jbd uhci_hcd ohci_hcd ehci_hcd
Pid: 16126, comm: ls_15964 Tainted: G      2.6.18-92.1.1.el5 #1
RIP: 0010:[<ffffffff80064b54>]  [<ffffffff80064b54>] .text.lock.spinlock+0x2/0x30
RSP: 0018:ffff81001d279dd0  EFLAGS: 00000286
RAX: 0000000000003efe RBX: ffff8100104b2900 RCX: 0000000000000000
RDX: ffff8100104b2900 RSI: ffff8100188ee860 RDI: ffffffff883f64c0
RBP: ffffffff80061e22 R08: ffff81001d278000 R09: 0000000000000001
R10: ffff810016f059c8 R11: 0000000300000000 R12: 0000000000000282
R13: a3754e699d64d231 R14: ffff81000563c000 R15: 0000000000003efe
FS:  0000000000000000(0000) GS:ffffffff8039f000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00007fff904eeaf9 CR3: 000000000dff2000 CR4: 00000000000006e0

Call Trace:
 [<ffffffff883e137f>] :stap_7c70ff0a5aa33e45a459d750475957f4_763:usr_itrace_init+0x92/0x17d
 [<ffffffff883e51e8>] :stap_7c70ff0a5aa33e45a459d750475957f4_763:__stp_utrace_attach_match_tsk+0x17c/0x22e
 [<ffffffff883e52e7>] :stap_7c70ff0a5aa33e45a459d750475957f4_763:__stp_utrace_task_finder_report_exec+0x4d/0x4f
 [<ffffffff800bd589>] utrace_report_exec+0x58/0x187
 [<ffffffff8003f49b>] search_binary_handler+0x171/0x26d
 [<ffffffff8003e950>] do_execve+0x18c/0x243
 [<ffffffff80054760>] sys_execve+0x36/0x4c


This could have been caused by create_itrace_info's use of an
uninitialized spin_lock.  It is not clear that a spinlock is
appropriate at all, considering which of the protected calls could
block.


- FChE

Re: itrace patches

[Dave Nomura]

Frank Ch. Eigler wrote:
> Dave Nomura <dcnltc@us.ibm.com> writes:
>
>   
>> Are the two patches that I submitted OK to commit?
>> [...]
>>     
>
> Thanks for committing.  I'll move the test suite file to the
> appropriate place.  While testing that though, I found a problem, so
>   
I've been on vacation for the last few weeks and will try to look into 
this today.

I've run the testsuite on ppc64/RHEL5.2, and I think I ran it on x86/SLES10
> I'll initially disable the test.  Can you elaborate upon the level of
> testing of this code? 

>  Has it been run under lockdep?
>   
No.  Can you elaborate on how to do this?
>
> On an RHEL5 x86-64 machine, I got:
>
> BUG: soft lockup - CPU#0 stuck for 10s! [ls_15964:16126]
> CPU 0:
> Modules linked in: stap_7c70ff0a5aa33e45a459d750475957f4_763(U) xt_tcpudp iptable_nat ip_nat ip_conntrack nfnet
> link ip_tables x_tables xfrm_nalgo crypto_api hidp l2cap bluetooth nfs lockd fscache nfs_acl sunrpc dm_multipat
> h video sbs backlight i2c_ec button battery asus_acpi acpi_memhotplug ac lp floppy ata_piix libata scsi_mod i2c
> _piix4 i2c_core pcspkr ide_cd 8139too parport_pc 8139cp mii parport serio_raw cdrom dm_snapshot dm_zero dm_mirr
> or dm_mod ext3 jbd uhci_hcd ohci_hcd ehci_hcd
> Pid: 16126, comm: ls_15964 Tainted: G      2.6.18-92.1.1.el5 #1
> RIP: 0010:[<ffffffff80064b54>]  [<ffffffff80064b54>] .text.lock.spinlock+0x2/0x30
> RSP: 0018:ffff81001d279dd0  EFLAGS: 00000286
> RAX: 0000000000003efe RBX: ffff8100104b2900 RCX: 0000000000000000
> RDX: ffff8100104b2900 RSI: ffff8100188ee860 RDI: ffffffff883f64c0
> RBP: ffffffff80061e22 R08: ffff81001d278000 R09: 0000000000000001
> R10: ffff810016f059c8 R11: 0000000300000000 R12: 0000000000000282
> R13: a3754e699d64d231 R14: ffff81000563c000 R15: 0000000000003efe
> FS:  0000000000000000(0000) GS:ffffffff8039f000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> CR2: 00007fff904eeaf9 CR3: 000000000dff2000 CR4: 00000000000006e0
>
> Call Trace:
>  [<ffffffff883e137f>] :stap_7c70ff0a5aa33e45a459d750475957f4_763:usr_itrace_init+0x92/0x17d
>  [<ffffffff883e51e8>] :stap_7c70ff0a5aa33e45a459d750475957f4_763:__stp_utrace_attach_match_tsk+0x17c/0x22e
>  [<ffffffff883e52e7>] :stap_7c70ff0a5aa33e45a459d750475957f4_763:__stp_utrace_task_finder_report_exec+0x4d/0x4f
>  [<ffffffff800bd589>] utrace_report_exec+0x58/0x187
>  [<ffffffff8003f49b>] search_binary_handler+0x171/0x26d
>  [<ffffffff8003e950>] do_execve+0x18c/0x243
>  [<ffffffff80054760>] sys_execve+0x36/0x4c
>
>
> This could have been caused by create_itrace_info's use of an
> uninitialized spin_lock.  It is not clear that a spinlock is
> appropriate at all, considering which of the protected calls could
> block.
>
>
> - FChE
>   


-- 
Dave Nomura
LTC Linux Power Toolchain

Re: itrace patches

[Frank Ch. Eigler]

Hi -

On Mon, Aug 25, 2008 at 08:02:59AM -0700, Dave Nomura wrote:
> [...]
> > Has [itrace] been run under lockdep?
> No.  Can you elaborate on how to do this?

You need to find or make a kernel that has CONFIG_LOCKDEP=y turned on.
Fedora makes these available as "kernel-debug-..." builds (not to
be confused with the kernel-*-debuginfo packages).

- FChE

Re: itrace patches

[Maynard Johnson]

Hi, Frank,
As I indicated in my previous posting, Dave has moved on to another position, so 
I'm taking over this work.  I understand that as a contributor, Dave also had 
commit rights to commit his own code (once approved by you).  Can you transfer 
that ability to me?

Thanks.
-Maynard Johnson