[Intel-gfx] SEGV w/ UXA and x11perf...

[Daniel J Blueman]

When gathering performance comparisons with kernel CONFIG_X86_PAT
disabled and enabled for Intel, radeon and nouveau drivers (during
Ubuntu 9.10 pre-release tuning), I found it's readily possible to
crash X with Intel drivers:

$ x11perf -shmputxy500

Setup is Intel 2.9.0, Xorg 1.6.4 on GM45, 2.6.31.5 kernel x86-64.

Maybe it's worthwhile adding 'x11perf -all -repeat 1 -reps 3' to any
regression testsuite?

Let me know if it helps if I log this into a freedesktop.org bug
report or otherwise.

Thanks,
  Daniel

--- [1]

Program terminated with signal 11, Segmentation fault.
#0  _mm_loadu_si128 (src_bits=<value optimised out>, dst_bits=<value
optimised out>, src_stride=2400, dst_stride=6144, src_bpp=<value
optimised out>,
    dst_bpp=<value optimised out>, src_x=<value optimised out>,
src_y=576, dst_x=<value optimised out>, dst_y=3, width=10, height=10)
    at /usr/lib/gcc/x86_64-linux-gnu/4.4.0/include/emmintrin.h:685

warning: Source file is more recent than executable.
685	  return (__m128i) __builtin_ia32_loaddqu ((char const *)__P);
(gdb) where
#0  _mm_loadu_si128 (src_bits=<value optimised out>, dst_bits=<value
optimised out>, src_stride=2400, dst_stride=6144, src_bpp=<value
optimised out>,
    dst_bpp=<value optimised out>, src_x=<value optimised out>,
src_y=576, dst_x=<value optimised out>, dst_y=3, width=10, height=10)
    at /usr/lib/gcc/x86_64-linux-gnu/4.4.0/include/emmintrin.h:685
#1  load128Unaligned (src_bits=<value optimised out>, dst_bits=<value
optimised out>, src_stride=2400, dst_stride=6144, src_bpp=<value
optimised out>,
    dst_bpp=<value optimised out>, src_x=<value optimised out>,
src_y=576, dst_x=<value optimised out>, dst_y=3, width=10, height=10)
at ../../pixman/pixman-sse2.c:316
#2  pixmanBltsse2 (src_bits=<value optimised out>, dst_bits=<value
optimised out>, src_stride=2400, dst_stride=6144, src_bpp=<value
optimised out>,
    dst_bpp=<value optimised out>, src_x=<value optimised out>,
src_y=576, dst_x=<value optimised out>, dst_y=3, width=10, height=10)
at ../../pixman/pixman-sse2.c:4518
#3  0x00007f3cd691132f in fbCopyNtoN (pSrcDrawable=<value optimised
out>, pDstDrawable=<value optimised out>, pGC=<value optimised out>,
pbox=0x7fff321f23f0, nbox=0,
    dx=<value optimised out>, dy=573, reverse=0, upsidedown=0,
bitplane=0, closure=0x0) at ../../fb/fbcopy.c:64
#4  0x00007f3cd6d88980 in uxa_copy_n_to_n (pSrcDrawable=<value
optimised out>, pDstDrawable=<value optimised out>, pGC=<value
optimised out>, pbox=0x3b92740,
    nbox=<value optimised out>, dx=573, dy=573, reverse=0,
upsidedown=0, bitplane=0, closure=0x0) at ../../uxa/uxa-accel.c:484
#5  0x00007f3cd691036d in fbCopyRegion (pSrcDrawable=0x3a311a0,
pDstDrawable=0x10, pGC=0x7f3cd4891010, pDstRegion=<value optimised
out>, dx=573, dy=<value optimised out>,
    copyProc=0x7f3cd6d88290 <uxa_copy_n_to_n>, bitPlane=0,
closure=0x5) at ../../fb/fbcopy.c:396
#6  0x00007f3cd691087a in fbDoCopy (pSrcDrawable=0x3a311a0,
pDstDrawable=0x39f6320, pGC=0x3a309b0, xIn=576, yIn=576,
widthSrc=<value optimised out>, heightSrc=10, xOut=0,
    yOut=0, copyProc=0x7f3cd6d88290 <uxa_copy_n_to_n>, bitPlane=0,
closure=0x0) at ../../fb/fbcopy.c:596
#7  0x00007f3cd6d88210 in uxa_copy_area (pSrcDrawable=0x3a311a0,
pDstDrawable=0x39f6320, pGC=0x3a309b0, srcx=576, srcy=<value optimised
out>, width=<value optimised out>,
    height=10, dstx=0, dsty=0) at ../../uxa/uxa-accel.c:503
#8  0x000000000053a248 in damageCopyArea (pSrc=0x3a311a0,
pDst=0x39f6320, pGC=0x3a309b0, srcx=<value optimised out>, srcy=<value
optimised out>, width=10, height=10, dstx=0,
    dsty=0) at ../../../miext/damage/damage.c:949
#9  0x000000000050e4c9 in doShmPutImage (client=0x3a324e0) at
../../Xext/shm.c:509
#10 ProcShmPutImage (client=0x3a324e0) at ../../Xext/shm.c:865
#11 0x000000000050f450 in ProcShmDispatch (client=0x3a324e0) at
../../Xext/shm.c:1119
#12 0x000000000044e174 in Dispatch () at ../../dix/dispatch.c:456
#13 0x0000000000434085 in main (argc=<value optimised out>,
argv=0x7fff321f2978, envp=<value optimised out>) at
../../dix/main.c:397
-- 
Daniel J Blueman