On Wed, 7 Apr 2004, Ulrich Teichert wrote:
> I'm having a wierd problem with XFree86 4.3.0.2 (build from source on
> x86-linux, running on a settop box with a Geode chip, nsc driver, keyboard
> and mouse over TV-IR and fed into /dev/tty0 and /dev/gpmdata, repectively)
> First, the X server runs like a charm. Stable and reliable, does not seem
> to have memory leaks or something. Very nice, until it comes to stopping
> it. CTRL-C on the console or SIGTERM does not terminate the server and
> it is using all CPU it can get in a busy loop. It is killable with SIGKILL,
> though.
> I really would like to understand the problem behind it, before I start
> fiddeling with a workaround (like killing the Xserver with SIGKILL instead
> of SIGTERM anytime). I suspect it's just a configuration issue.
[...]
> I traced it down where the server is hanging, it's the same spot everytime,
> but I wasn't able to see the C-source of it, despite building *everything*
> with debug and using a libc with debugging symbols (nfs mounted - the box
> does not have a hard disk). The stack trace is:
> #0 0x401a875c in ?? ()
> #1 0x4018a6b7 in ?? ()
> #2 0x4018a6e6 in ?? ()
> #3 0x081667da in AnimCurCloseScreen (index=0, pScreen=0x83327e0)
> at animcur.c:126
> #4 0x080c07c5 in main (argc=8, argv=0xbffffdc4, envp=0xbffffde8) at main.c:459
> Line 126 in animcur.c is:
> ret = (*pScreen->CloseScreen) (index, pScreen);
> (gdb) p screenInfo
> $2 = {imageByteOrder = 0, bitmapScanlineUnit = 32, bitmapScanlinePad = 32,
> bitmapBitOrder = 0, numPixmapFormats = 7, formats = {{depth = 1 '\001',
> bitsPerPixel = 1 '\001', scanlinePad = 32 ' '}, {depth = 4 '\004',
> bitsPerPixel = 8 '\b', scanlinePad = 32 ' '}, {depth = 8 '\b',
> bitsPerPixel = 8 '\b', scanlinePad = 32 ' '}, {depth = 15 '\017',
> bitsPerPixel = 16 '\020', scanlinePad = 32 ' '}, {depth = 16 '\020',
> bitsPerPixel = 16 '\020', scanlinePad = 32 ' '}, {depth = 24 '\030',
> bitsPerPixel = 32 ' ', scanlinePad = 32 ' '}, {depth = 32 ' ',
> bitsPerPixel = 32 ' ', scanlinePad = 32 ' '}, {depth = 0 '\0',
> bitsPerPixel = 0 '\0', scanlinePad = 0 '\0'}}, arraySize = 16,
> numScreens = 1, screens = {0x83327e0, 0x0 <repeats 15 times>},
> numVideoScreens = 1}
> (gdb) p pScreen
> $6 = 0x3d5
> The disassembled x86 code where it loops is:
> 0x401a9759: mov %ebx,%edx
> 0x401a975b: in (%dx),%al
> 0x401a975c: test %al,%al
> 0x401a975e: jl 0x401a9759
> 0x401a9760: mov $0x30,%al
> 0x401a9762: mov %esi,%edx
> 0x401a9764: out %al,(%dx)
> (gdb) info all-registers
> eax 0xff 255
> ecx 0x3d5 981
> edx 0x3d5 981
> ebx 0x3d5 981
> [snip]
> The loop starts at 0x401a9759 and goes to 0x401a975e. If I jump to
> 0x401a9760 in the debugger, the server terminates. So, it reads 0x3d5,
> whatever that is (probably I/O stuff) and the test never succeeds.
Looks like it's looping in function gfx_vga_mode_switch() near the tail end of
xc/programs/Xserver/hw/xfree86/drivers/ncs/gfx/vga_gu1.c, part of which reads
/* WAIT UNTIL SOFTVGA HAS VALIDATED MODE IF ENDING MODE SWITCH */
/* This is for VSA1 only, where SoftVGA waits until the next */
/* vertical blank to validate the hardware state. */
if ((!active) && (!(gu1_detect_vsa2()))) {
OUTB(crtcindex, 0x33);
while (INB(crtcdata) & 0x80) ;
}
Change this to ...
/* WAIT UNTIL SOFTVGA HAS VALIDATED MODE IF ENDING MODE SWITCH */
/* This is for VSA1 only, where SoftVGA waits until the next */
/* vertical blank to validate the hardware state. */
if ((!active) && (!(gu1_detect_vsa2()))) {
int cnt = 256;
OUTB(crtcindex, 0x33);
while (--cnt && (INB(crtcdata) & 0x80));
}
... and see what happens.
Marc.
+----------------------------------+-----------------------------------+
| Marc Aurele La France | work: 1-780-492-9310 |
| Computing and Network Services | fax: 1-780-492-1729 |
| 352 General Services Building | email: tsi@xxxxxxxxxxx |
| University of Alberta +-----------------------------------+
| Edmonton, Alberta | |
| T6G 2H1 | Standard disclaimers apply |
| CANADA | |
+----------------------------------+-----------------------------------+
XFree86 developer and VP. ATI driver and X server internals.
_______________________________________________
XFree86 mailing list
XFree86@xxxxxxxxxxx
http://XFree86.Org/mailman/listinfo/xfree86
