On 24/06/16 17:37, Chris Wilson wrote:
On Fri, Jun 24, 2016 at 05:04:46PM +0100, Matthew Auld wrote:
The gen6_for_all_pdes macro does the upper-bound evaluation after
accessing the page_table array, hence on the final iteration we end up
hitting an out-of-bounds error:
[ 1023.831657] UBSAN: Undefined behaviour in drivers/gpu/drm/i915/i915_gem_gtt.c:1993:2
[ 1023.831680] index 512 is out of range for type 'i915_page_table *[512]'
[ 1023.831696] CPU: 0 PID: 4833 Comm: rmmod Tainted: G U 4.7.0-rc4-drm-intel-debug+ #5
[ 1023.831698] Hardware name: ASUS All Series/Z87-K, BIOS 1202 05/13/2014
[ 1023.831700] 0000000000000200 00000000adfe9733 ffff8801a3917988 ffffffff818cc0a4
[ 1023.831705] 0000000041b58ab3 ffffffff8275ca08 ffffffff818cbff2 ffff8801a39179b0
[ 1023.831708] ffff8801a3917960 0000000000000200 1ffffffff4365b17 0000000000000001
[ 1023.831711] Call Trace:
[ 1023.831717] [<ffffffff818cc0a4>] dump_stack+0xb2/0x10e
[ 1023.831721] [<ffffffff818cbff2>] ? _atomic_dec_and_lock+0x152/0x152
[ 1023.831726] [<ffffffff81952b0b>] ubsan_epilogue+0xd/0x4e
[ 1023.831730] [<ffffffff8195373d>] __ubsan_handle_out_of_bounds+0x107/0x14d
[ 1023.831733] [<ffffffff81953636>] ? __ubsan_handle_shift_out_of_bounds+0x24c/0x24c
[ 1023.831737] [<ffffffff814bfde6>] ? kfree+0x246/0x3f0
[ 1023.831801] [<ffffffffa183bff8>] gen6_ppgtt_cleanup+0x128/0x130 [i915]
Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx>
Ok. Tried to find something to complain about and couldn't.
Reviewed-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
-Chris
Well ... not enough to reject it, but there's the lack of parentheses
round macro parameters, and it uses ?: rather than the && style used in
the Gen8 equivalents. I'll post an alternative based on the Gen8 version ...
.Dave.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
