Occasionally I have seen something similar, mostly when I've just introduced a bug into the driver, but nevertheless.

It's basically an integer overflow (resulting in an invalid access) when trying to draw an out-of-bounds rectangle. I've fixed up a couple of that caused hangs, so a slightly different code path.

Created attachment 36139 [details] [review]
Trim fills to drawable bounds.

Michael if you know of a way to reproduce the crash and could test this patch, that would be excellent.

Hi Chris,

seems like this problem arose when setting very large window coordinates (due to a bug I was setting (2, 0, -8, -23) as window rect while the latter two are interpreted as unsigned values of course).

After applying your patch, Xorg does not crash any longer. Thanks!

Best regards,
Michael

After applying the patch I sometimes run into a (nearly?) endless loop in Xorg, for example when opening xpdf. Can you see if your patch is causing this? Backtrace comes here:

0  fbBltOne (src=0x0, srcStride=<value optimized out>, srcX=<value optimized out>, dst=0x9869dba14928,
    dstStride=<value optimized out>, dstX=<value optimized out>, dstBpp=32, width=160, height=-32, fgand=0, fgxor=0,
    bgand=4294967295, bgxor=0) at ../../fb/fbbltone.c:418
#1  0x00007fe5b29cdb80 in fbOddStipple (dst=<value optimized out>, dstStride=<value optimized out>,
    dstX=<value optimized out>, dstBpp=32, width=<value optimized out>, height=<value optimized out>, stip=0x7fe5b1bdc000,
    stipStride=16, stipWidth=5, stipHeight=14, fgand=0, fgxor=0, bgand=4294967295, bgxor=0, xRot=69, yRot=0)
    at ../../fb/fbstipple.c:261
#2  0x00007fe5b29c57eb in fbFill (pDrawable=<value optimized out>, pGC=0x2e3e050, x=<value optimized out>,
    y=<value optimized out>, width=5, height=-32) at ../../fb/fbfill.c:133
#3  0x00007fe5b29c5d0a in fbPolyFillRect (pDrawable=<value optimized out>, pGC=<value optimized out>,
    nrect=<value optimized out>, prect=<value optimized out>) at ../../fb/fbfillrect.c:77
#4  0x00007fe5b2e43011 in uxa_check_poly_fill_rect (pDrawable=0x28124b0, pGC=0x2e3e050, nrect=1, prect=0x27b2588)
    at ../../uxa/uxa-unaccel.c:257
#5  0x00007fe5b2e3ccf0 in uxa_poly_fill_rect (pDrawable=0x28124b0, pGC=0x2e3e050, nrect=1, prect=0x27b2588)
    at ../../uxa/uxa-accel.c:727
#6  0x00000000004c329b in damagePolyFillRect (pDrawable=0x28124b0, pGC=0x2e3e050, nRects=1, pRects=0x27b2588)
    at ../../../miext/damage/damage.c:1404
#7  0x0000000000439e24 in ProcPolyFillRectangle (client=0x2b35cb0) at ../../dix/dispatch.c:1939
#8  0x000000000043c9a4 in Dispatch () at ../../dix/dispatch.c:439
#9  0x0000000000425b4a in main (argc=7, argv=0x7d7c88, envp=<value optimized out>) at ../../dix/main.c:285

As far as I know that is a separate bug, caused by a page-fault-of-doom.

Found the bug reference I was looking for... bug 28478 for the endless loop.

Even though I am using your patch, I just got another crash when running xxdiff. Backtrace comes here:

(gdb) bt full
#0  0x00007fb06fff5175 in *__GI_raise (sig=<value optimized out>) at ../nptl/sysdeps/unix/sysv/linux/raise.c:64
        pid = <value optimized out>
        selftid = <value optimized out>
#1  0x00007fb06fff7f80 in *__GI_abort () at abort.c:92
        act = {__sigaction_handler = {sa_handler = 0, sa_sigaction = 0}, sa_mask = {__val = {0, 3, 0, 0, 0, 140395769797384, 140395802543288, 
              0, 4294967295, 1179670597, 1, 8102920, 0, 1073741823, 140395692755240, 0}}, sa_flags = 1909584850, 
          sa_restorer = 0x100041500000001}
        sigs = {__val = {32, 0 <repeats 15 times>}}
#2  0x000000000047d783 in ddxGiveUp () at ../../../../hw/xfree86/common/xf86Init.c:1214
        i = <value optimized out>
#3  0x000000000046368d in AbortServer () at ../../os/log.c:404
No locals.
#4  0x0000000000463d2e in FatalError (f=0x5734a0 "Caught signal %d (%s). Server aborting\n") at ../../os/log.c:529
        args = {{gp_offset = 24, fp_offset = 48, overflow_arg_area = 0x7fffb8ed2720, reg_save_area = 0x7fffb8ed2660}}
        beenhere = 1
#5  0x0000000000467f3e in OsSigHandler (signo=11, sip=0x7fb06b650928, unused=<value optimized out>) at ../../os/osinit.c:156
No locals.
#6  <signal handler called>
No symbol table info available.
#7  fbBltOne (src=<value optimized out>, srcStride=<value optimized out>, srcX=<value optimized out>, dst=0x7fb06b650928, 
    dstStride=<value optimized out>, dstX=<value optimized out>, dstBpp=32, width=-832, height=-5, fgand=0, fgxor=4287072135, bgand=4294967295, 
    bgxor=0) at ../../fb/fbbltone.c:352
        fbBits = 0x7fb06db78790
        srcEnd = 0x7fb06d78a240
        pixelsPerDst = <value optimized out>
        leftShift = 2
        rightShift = 30
        startmask = 0
        endmask = 0
        bits = 1073741823
        bitsLeft = 0
        bitsRight = <value optimized out>
        left = <value optimized out>
        mask = <value optimized out>
        nDst = -26
        w = 0
        n = -26
        nmiddle = <value optimized out>
        copy = 0
        transparent = 1
        srcinc = <value optimized out>
        endNeedsLoad = 0
        fbLane = 0x7fb06dd7baf0 ""
        startbyte = <value optimized out>
        endbyte = 0
#8  0x00007fb06db76b80 in fbOddStipple (dst=<value optimized out>, dstStride=<value optimized out>, dstX=<value optimized out>, dstBpp=32, 
    width=<value optimized out>, height=<value optimized out>, stip=0x7fb06d78a000, stipStride=16, stipWidth=16, stipHeight=16, fgand=0, 
    fgxor=4287072135, bgand=4294967295, bgxor=0, xRot=584, yRot=-478) at ../../fb/fbstipple.c:261
        stipX = 2
        stipY = <value optimized out>
        sx = 1840698024
        widthTmp = 0
        h = -5
        w = -832
        x = 17920
---Type <return> to continue, or q <return> to quit---
        y = 0
#9  0x00007fb06db6e7eb in fbFill (pDrawable=<value optimized out>, pGC=0x15c8ec0, x=<value optimized out>, y=<value optimized out>, width=-26, 
    height=-5) at ../../fb/fbfill.c:133
        fgand = 0
        fgxor = 4287072135
        bgand = <value optimized out>
        bgxor = <value optimized out>
        pStip = <value optimized out>
        stipWidth = 16
        stipHeight = 16
        dst = 0x7fb06a638000
        dstStride = 2560
        dstBpp = 32
        dstXoff = 0
        dstYoff = <value optimized out>
        pPriv = 0x15c9080
        x1 = -26
        x2 = <value optimized out>
        y1 = <value optimized out>
        y2 = <value optimized out>
#10 0x00007fb06db6ed0a in fbPolyFillRect (pDrawable=<value optimized out>, pGC=<value optimized out>, nrect=<value optimized out>, 
    prect=<value optimized out>) at ../../fb/fbfillrect.c:77
        pClip = 0x2a848b0
        pbox = <value optimized out>
        extentX1 = 586
        extentX2 = 1142
        extentY1 = 82
        extentY2 = 631
        fullX1 = 586
        fullX2 = 1142
        fullY1 = 558
        fullY2 = 572
        partX1 = <value optimized out>
        partX2 = <value optimized out>
        partY1 = <value optimized out>
        partY2 = <value optimized out>
        xorg = 584
        yorg = 80
#11 0x00007fb06dfec011 in uxa_check_poly_fill_rect (pDrawable=0x35629a0, pGC=0x15c8ec0, nrect=1, prect=0x37ae988) at ../../uxa/uxa-unaccel.c:257
        __FUNCTION__ = "uxa_check_poly_fill_rect"
#12 0x00007fb06dfe5cf0 in uxa_poly_fill_rect (pDrawable=0x35629a0, pGC=0x15c8ec0, nrect=1, prect=0x37ae988) at ../../uxa/uxa-accel.c:727
        pClip = 0x2a848b0
        pPixmap = 0x252d000
        pbox = <value optimized out>
        extentX1 = 0
        extentX2 = 44589488
        extentY1 = 32688
        extentY2 = 1842878571
        fullX1 = <value optimized out>
        fullX2 = <value optimized out>
        fullY1 = <value optimized out>
        fullY2 = <value optimized out>
        partX1 = <value optimized out>
        partX2 = <value optimized out>
        partY1 = <value optimized out>
---Type <return> to continue, or q <return> to quit---
        partY2 = <value optimized out>
        xoff = <value optimized out>
        yoff = <value optimized out>
        xorg = 0
        yorg = 44589488
        pReg = 0x3d36f10
#13 0x00000000004c329b in damagePolyFillRect (pDrawable=0x35629a0, pGC=0x15c8ec0, nRects=1, pRects=0x37ae988)
    at ../../../miext/damage/damage.c:1404
        pGCPriv = 0x275b2b0
        oldFuncs = 0x7c3ae0
#14 0x0000000000439e24 in ProcPolyFillRectangle (client=0x18107d0) at ../../dix/dispatch.c:1939
        things = 1840698024
        pGC = 0x15c8ec0
        pDraw = 0x35629a0
#15 0x000000000043c9a4 in Dispatch () at ../../dix/dispatch.c:439
        result = <value optimized out>
        client = 0x18107d0
        nready = 0
        start_tick = 167320
#16 0x0000000000425b4a in main (argc=7, argv=0x7d7c88, envp=<value optimized out>) at ../../dix/main.c:285
        i = 1
        alwaysCheckForInput = {0, 1}

In a sense, that is a good thing since the patch *should* have had no
effect and appeared to be a band-aid for a deeper bug. This means that I
need to keep digging. Thanks.

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/xorg/driver/xf86-video-intel/issues/5.