On Wed, 3 Nov 2004, kurt emmerich wrote: > Thanks Mark, > > The problem is exhibited on two systems- a desktop and a laptop, both running fedora core 1. The desktop has been updated to fedora core 2 and also exhibits the problem. > I am at the moment updating the desktop with the latest fc2 updates. > The fedora 1 kernel is 2.4.22-1.2115 > > 1) Dell desktop 400MHz Pentium II > video card S3 ViRGE [PCI slot] (no card installed in AGP slot) > sound card Ensoniq ES1371 [AudioPCI-97] on-board > generic ps2 mouse > 3COM 3c905B 100Baase TX [Cyclone] network card [PCI slot] > 2 com ports on-board > 1 parallel port > usb host on board > 256Meg ram > hard drive/cd rom/floppy Make sure the ViRGE driver does not have the pci_retry option enabled (see the s3virge man page). You should also try Option "NoAccel" to make sure it's not something else related to the way the ViRGE driver is programming the hardware. > > 2) Dell Inspiron 5000 laptop 500MHz Pentium III > video ATI Rage Mobility P/M AGP 2X > > ethernet, pcmcia > 1 serial port > usb host > xxx meg ram > hard drive/cd rom/floppy > ps2 mouse/ integrated keyboard etc > > On the desktop: > > I believe a frame buffer is not being used ? /var/log/XFree86.0.log indicates (in no particular order) (FC2 uses /var/log/Xorg.o.log and indicates similar entries) > S3VIRGE(0): Using fb > (II) Loading sub module "fb" > (II) LoadModule: "fb" > (II) Loading /usr/X11R6/lib/modules/libfb.a > (II) Module fb: vendor ="The XFree86 Project" > LoadModule: "fbdevhw" > The 'Section "Device"' entry does not contain a MemBase or IOBase entry. If the fbdevhw module is being loaded explictily in the Section "Module" of the XF86Config file you should remove it. That module would get autoloaded only if it was needed (same for most modules). But it looks like the ViRGE driver doesn't support using the framebuffer device. The ATI driver supports a UseFBDev option, however. > > running lspci -v, references are made to > 00:00.0 Host Bridge: Intel Corp. 440/BX/ZX/DX - 82443BX/ZX/DX Host Bridge etc > (dmesg lists agpgart: Detected an Intel 440BX chipset) BX is a good chipset. It's the earlier LX model that was particularly susceptible to retry problems. > > reference is also made to ViRGE (desktop, the entry on the laptop is similar, uses irq 11) : > 00:10.0 VGA compatible controller S3 Inc ViRGE/DX or /GX (rev 01) (prog-iff 00 [VGA]) > subsystem: S3 Inc. ViRGE/DX > flags: bus master, medium devsel, latency 64, IRQ 10 > Memory at f8000000 (32-bit, non-prefetchable) [size=febf0000] > Expansion ROM at 00010000 [disabled] > > /dev/ttyS0 is using irq 4, /dev/ttyS1 is using irq 3. There is no sharing of the interrupt. > > There is no reference to "PCIRetry" in /etc/X11/XF86Config. This file does list the video "driver" as "s3virge". ( laptop lists "driver" "ati") in the "Device" section. > > I have searched diligently for some reference to this in google, but can find none. I find it hard to believe that someone else hasn't had the same difficulty ( or made the same mistakes if that is the case). > > btw, the serial ports are 16550A. I changed the receive fifo high-water-mark to 1, but saw no difference. > > > Most perplexing, is there anything I can look at? I think that if you're not using a framebuffer device and if you disable acceleration (both drivers support Option "NoAccel") and you still have the problem, the X-server is probably off the hook. The only things I could think of is interrupt problems (like shared interrupts) or bus-hogging problems (like pci retries). I'm not sure what KDE or Gnome could be doing to cause the problem, but you might try seeing if it happens with something simpler like twm. I guess one other thing to check is that /proc/interrupts doesn't show alot of interrupts for the graphics card. I don't think it should be generating interrupts. Mark. > > Thanks, > Regards, > Kurt > > > > Mark Vojkovich <mvojkovi@xxxxxxxxxxx> wrote: > On Tue, 2 Nov 2004, kurt emmerich wrote: > > > I hope this is the right place to ask - if not I apologize- would someone please point me in the right direction? > > > > I'm running Fedora 1 on a 400MHz Pentium 2. Runs great, no problem, except when using the KDE or GNOME desktop. When receiving serial data ( /dev/ttyS0) 115200,8,n,1, I'm losing characters. Looks to me like the serial receive interrupt is getting 'stalled'. This occurs at bauds as low as 19200 ( I dont recall if I tried lower than that) and gets must worse as I drag windows aroung the screen with the mouse. > > However, when I open a text console ( ctrl-alt-f1) and run minicom from there, I can successfully receive well over 150M of data with zero errors even at 115K baud. Hardware handshaking is enabled and asserted. > > > > Might I ask for feedback on this? Has anyone else seen this? Any opinions on why this is occurring?? > > > > Thanks, > > Regards, > > Kurt > > Which graphics card and driver? Which AGP chipset (ie. LX, BX, etc...)? > Your XFree86 driver isn't using a framebuffer device is it? > Is the serial port sharing an interrupt with something else? > > I suspect a shared interrupt or a "PCI retry" problem. PCI retries > are usually caused by the graphics driver and are particularly a problem > on Intel LX chipsets. Historically, this has caused problems with > sound cards, but I suppose it could effect serial as well. > > Mark. > _______________________________________________ > XFree86 mailing list > XFree86@xxxxxxxxxxx > http://XFree86.Org/mailman/listinfo/xfree86 > > > > > > > --------------------------------- > Do you Yahoo!? > Check out the new Yahoo! Front Page. www.yahoo.com/a _______________________________________________ XFree86 mailing list XFree86@xxxxxxxxxxx http://XFree86.Org/mailman/listinfo/xfree86
