[OOPS][2.6.0][e100 new driver] mii-diag oops with -F option

[OOPS][2.6.0][e100 new driver] mii-diag oops with -F option

[Shawn Starr]

Sorry, this is with the new Intel e100 pro driver not the old eepro100
driver.

Here is the oops dump below this seems ugly:

 Unable to handle kernel paging request at virtual address c046a640
  printing eip:
 c046a640
 *pde = 00103027
 *pte = 0046a000
 Oops: 0000 [#1]
 CPU:    0
 EIP:    0060:[<c046a640>]    Not tainted
 EFLAGS: 00010286
 EIP is at e100_phy_init+0x0/0x80
 eax: 00000000   ebx: e3593a40   ecx: 00000000   edx: e0402000
 esi: e3593a40   edi: 00000000   ebp: e35937f8   esp: e0403e8c
 ds: 007b   es: 007b   ss: 0068
 Process mii-diag (pid: 165, threadinfo=e0402000 task=e0453960)
 Stack: c0279763 e3593a40 e0403ea0 c0120d4a e35937f8 e35937f8 c027ee63 e3593a40
        40015035 e0402000 c0238910 e17d1000 00008949 00008949 ffffffed 00008949
        c033d698 e35937f8 e0403f0c 00008949 c04164a0 e0403f0c 00008949 00000001
 Call Trace:
  [<c0279763>] e100_hw_init+0x13/0x170
  [<c0120d4a>] preempt_schedule+0x2a/0x50
  [<c027ee63>] e100_mii_ioctl+0x1f3/0x320
  [<c0238910>] write_chan+0x170/0x230
  [<c033d698>] dev_ifsioc+0x3f8/0x470
  [<c033da0f>] dev_ioctl+0x2ff/0x3a0
  [<c0332540>] sock_ioctl+0x0/0x460
  [<c0389b3c>] inet_ioctl+0x11c/0x130
  [<c0332767>] sock_ioctl+0x227/0x460
  [<c01780b2>] vfs_write+0xd2/0x130
  [<c0192e19>] sys_ioctl+0x219/0x430
  [<c01781c2>] sys_write+0x42/0x70
  [<c010a41b>] syscall_call+0x7/0xb

 Code:  Bad EIP value.

The EIP matched in my System.map is e100_phy_init

Shawn.

Re: [OOPS][2.6.0][e100 new driver] mii-diag oops with -F option

[Jeff Garzik]

On Mon, Dec 22, 2003 at 12:37:49PM -0500, Shawn Starr wrote:
> Sorry, this is with the new Intel e100 pro driver not the old eepro100
> driver.

New is a bit non-specific :)  What driver version?

If you remove all occurences of "__devinit", does the oops go away?

RE: [OOPS][2.6.0][e100 new driver] mii-diag oops with -F option

[Feldman, Scott]

> New is a bit non-specific :)  What driver version?

It's got to be the e100-2.3.x driver in 2.6.0.
 
> If you remove all occurences of "__devinit", does the oops go away?

That's the problem.  __devinit's where there shouldn't be __devinit's.

So e100-3.0.x is pending in -exp queue for 2.6.1; do we fix 2.6.0?

-scott

Re: [OOPS][2.6.0][e100 new driver] mii-diag oops with -F option

[Shawn Starr]

This is correct, when I removed the __devinit's the oops did not occur.

I see this has been fixed in akpm's 2.6.0-mm1 release :-)

Thanks,

Shawn.

List:     linux-kernel
Subject:  RE: [OOPS][2.6.0][e100 new driver] mii-diag oops with -F option
Date:     2003-12-22 19:33:08

> New is a bit non-specific :)  What driver version?

It's got to be the e100-2.3.x driver in 2.6.0.

> If you remove all occurences of "__devinit", does the oops go away?

That's the problem.  __devinit's where there shouldn't be __devinit's.

So e100-3.0.x is pending in -exp queue for 2.6.1; do we fix 2.6.0?

-scott

Re: [OOPS][2.6.0][e100 new driver] mii-diag oops with -F option

[Jeff Garzik]

Feldman, Scott wrote:
> So e100-3.0.x is pending in -exp queue for 2.6.1; do we fix 2.6.0?


I'll go ahead and expedite the __devinit removal to Linus for now

	Jeff