Re: Fw: Slab coruption and oops with 2.6.1-mm4

[Gerd Knorr]

> heh, this is the same bug.  Last time we were unlocking an unlocked page. 
> Now we're freeing a free page.

Yes.  Still no idea why that happens through ...

> CONFIG_PREEMPT=y

Bug reproducable with this one turned off?

> Slab corruption: start=c57c2000, len=4096
> 000: 6e 72 6d 71 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b
> bttv0: skipped frame. no signal? high irq latency? [main=b030000,o_vbi=b030018,o_field=5378000,rc=537801c]
> ------------[ cut here ]------------
> kernel BUG at include/linux/mm.h:275!

page_cache_release()

> EIP is at videobuf_dma_free+0xa9/0xc0 [video_buf]

The code calling page_cache_release looks like this ...

	if (dma->pages) {
		int i;
		for (i=0; i < dma->nr_pages; i++)
			page_cache_release(dma->pages[i]);
		kfree(dma->pages);
		dma->pages = NULL;
	}

... even with videobuf_dma_free() called twice by mistake that shouldn't
double-free the pages.  Maybe videobuf_dma_free() is called from two
places at the same time because one of the call paths misses a lock, but
I can't find any on a quick review.  Hmm.

Does transcode use threads?  If so, does it call into bttv from
different threads?

> Call Trace:
>  [<d08f3a70>] bttv_dma_free+0x60/0xa0 [bttv]
>  [<d08ede63>] bttv_do_ioctl+0x403/0x16a0 [bttv]

must be VIDIOCSYNC ioctl.

>  [<c0335498>] video_usercopy+0xe8/0x1e0
>  [<d08ef13e>] bttv_ioctl+0x3e/0x70 [bttv]
>  [<c0168ef3>] sys_ioctl+0xf3/0x280
>  [<c042e9b7>] syscall_call+0x7/0xb

  Gerd

-- 
"... und auch das ganze Wochenende oll" -- Wetterbericht auf RadioEins
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/