[lkml]   [2011]   [Feb]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Patch in this message
SubjectRe: Linux 2.6.38-rc4 (other bugs: ipmi Oops)
On Thu, Feb 10, 2011 at 11:34 AM, Randy Dunlap <> wrote:
> Loading ipmi_si module a second time causes an Oops:
> [   68.120143] RIP: 0010:[<ffffffff813fc579>]  [<ffffffff813fc579>] put_driver+0x10/0x22

The disassembly is

55 push %rbp
48 89 e5 mov %rsp,%rbp
0f 1f 44 00 00 nopl 0x0(%rax,%rax,1)
48 ff 05 c7 af 80 01 incq 0x180afc7(%rip) # 0x180aff2
* 48 8b 7f 60 mov 0x60(%rdi),%rdi <-- trapping instruction
e8 38 27 ec ff callq 0xffffffffffec276c
48 ff 05 bf af 80 01 incq 0x180afbf(%rip) # 0x180affa
c9 leaveq
c3 retq

which is the access of "drv->p" in that function:


so "drv" that was passed in was just bogus. (it's
"0xffffffffa06a8430", looks like it's the DEBUG_PAGEALLOC that has
caused the page to be free'd).

> [   68.340115] Call Trace:
> [   68.340115]  [<ffffffff813fc64b>] driver_register+0xc0/0x1b2
> [   68.340115]  [<ffffffff8137f5de>] pnp_register_driver+0x28/0x31
> [   68.340115]  [<ffffffffa06b888d>] init_ipmi_si+0x1a4/0x4cd [ipmi_si]
> [   68.340115]  [<ffffffff810020a6>] do_one_initcall+0x6c/0x1e3
> [   68.340115]  [<ffffffff810d4998>] sys_init_module+0x12b/0x307

And I think that - as usual - the problem is that the damn driver
cleanup is very ugly, and has this duplicate set of code to unregister
all the random crap. Except one of the duplicates is missing one case.
I think the bug was introduced by Gjorn Helgaas in commit 9e368fa011d4
("ipmi: add PNP discovery (ACPI namespace via PNPACPI)") which added
the acpi pnp case, but only unregistered it on the regular module exit
path, not on the "module loaded with no pnp devices" path.

Does this patch fix it? And Corey - this is a good example of why the
code shouldn't duplicate the "unregister stuff" in the module load
error case vs the module exit path, and there should be a shared
"cleanup()" function that is called by both. Can this be cleaned up,


drivers/char/ipmi/ipmi_si_intf.c | 4 ++++
1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
index b6ae6e9..5953825 100644
--- a/drivers/char/ipmi/ipmi_si_intf.c
+++ b/drivers/char/ipmi/ipmi_si_intf.c
@@ -3454,6 +3454,10 @@ static int __devinit init_ipmi_si(void)
if (pci_registered)
+ if (pnp_registered)
+ pnp_unregister_driver(&ipmi_pnp_driver);

if (of_registered)
 \ /
  Last update: 2011-02-10 21:07    [W:0.137 / U:6.240 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site