Distribution: Debian Woody with Bunk debs Hardware Environment: Acer TravelMate 242, latest BIOS 1.15 Software Environment: linux-vanilla-2.6 Problem Description: When starting pcmcia-cs utils kernel totally freezes. Using pci=noacpi makes it usable. 2.6.5 worked OK 2.4.26 works OK Steps to reproduce: use my .config and try to recompile and use
Created attachment 3143 [details] Config for 2.6
Could you please attach the dmesg, acpidmp, and lspci -vv output from 2.6.5. We want to check it. Thanks.
Hmm my friend reported the same problem on Dell Latitude D600.
At present time I need to recompile it, coz I deleted it. Sending dmesg, lspci -vv, /proc/cpuinfo from 2.4.26. It should be helpfull too.
Created attachment 3157 [details] cpuinfo_2.4.26
Created attachment 3158 [details] lspci_-vv_2.4.26
Created attachment 3159 [details] dmesg_2.4.26
Created attachment 3160 [details] dmesg_2.6.4
Created attachment 3161 [details] acpidmp_2.6.4
Changing affected kernels to 2.6.5 too
Could you please provide dmesg and lspci -vv output in failure situation as well? so we can compare them. I think one way is temporarily disable pcmcia- cs, so you can get these output. Thanks.
Hi, when I edit /etc/init.d/pcmcia and at the beginning I add exit 0, kernel freezes too when kernel looks for yenta socket. So only way is to remove yenta prom kernel config...do you want me to do it?
Well, I done it other way: in lilo: linux 2.6.6 init=/bin/sh remountet / rw dmesg and here it is :)
Created attachment 3162 [details] dmesg_2.6.6_init=/bin/sh
Created attachment 3163 [details] lspci_-vv_2.6.6
Great method to get dmesg :). Looks that card bus's IRQ have the same value (IRQ 10) in 2.6.4 and 2.6.6, I'm a little curious why pci=noacpi makes your system work. In freeze situation, did you leave any pc card in your card bus slot?
Thanks :) Yes, I have my Zcom xi-626 in slot, but I tested it without this card in slot too, but no change. Only disabling yenta in kernel helped.
When boot with pci=noacpi, this message starts to repeat in dmesg - about 2 per 10 seconds DMA write timed out parport0: FIFO is stuck parport0: BUSY timeout (1) in compat_write_block_pio
Hi, I upgraded to debian-testing and now kernel shows this before crash: List is taken from pci=noacpi NET: Registered protocol family 23 cs: IO port probe 0x0100-0x04ff: excluding 0x378-0x37f cs: IO port probe 0x0800-0x08ff: excluding 0x800-0x80f cs: IO port probe 0x0c00-0x0cff: clean. cs: IO port probe 0x0a00-0x0aff: clean. cs: memory probe 0xa0000000-0xa0ffffff: clean. hmm and maybe problem is in pnp?! --------------------------------- pnp: Device 00:16 activated. parport: PnPBIOS parport detected. parport0: PC-style at 0x278 (0x678), irq 5, dma 3 [PCSPP,TRISTATE,COMPAT,EPP,ECP,DMA] lp0: using parport0 (interrupt-driven). Bluetooth: L2CAP ver 2.2 Bluetooth: L2CAP socket layer initialized Bluetooth: RFCOMM ver 1.3 Bluetooth: RFCOMM socket layer initialized Bluetooth: RFCOMM TTY layer initialized mtrr: base(0xe8020000) is not aligned on a size(0x300000) boundary DMA write timed out Bluetooth: HIDP (Human Interface Emulation) ver 0.3
PNP is required, since it claims some resources for motherboard and avoid resources conflict. could you please attach the cat /proc/ioports and /proc/iomem output in 2.6.6 and 2.6.4, so I can check if there is resources conflict. can acpi=off makes this system work? Thanks.
Created attachment 3207 [details] proc_iomem_2.6.4
Created attachment 3208 [details] proc_ioports_2.6.4
Created attachment 3209 [details] proc_iomem_pci=noacpi_2.6.7
Created attachment 3210 [details] proc_ioports_pci=noacpi_2.6.7
Here they are. pci=noacpi helps to my system to boot.
Thanks for the info. did this system work with acpi=noirq? and could you please attach the iomem/ioport without pci=noacpi? (maybe using your method 'init=/bin/sh') I must compare success situation and failure situation. Thanks.
Created attachment 3211 [details] proc_iomem_init=/bin/sh_2.6.7
Created attachment 3212 [details] proc_ioports_init=/bin/sh_2.6.7
Created attachment 3213 [details] proc_iomem_acpi=noirq_2.6.7
Created attachment 3214 [details] proc_ioports_acpi=noirq_2.6.7
when acpi=noirq, system boots well too
Thanks. Hmm, acpi=noirq makes the system work. This suggest it's an IRQ problem. please apply below patch, and try: ===== drivers/acpi/pci_link.c 1.28 vs edited ===== --- 1.28/drivers/acpi/pci_link.c Tue May 11 04:42:35 2004 +++ edited/drivers/acpi/pci_link.c Mon Jun 21 16:28:10 2004 @@ -553,7 +553,7 @@ if (acpi_strict) printk(KERN_WARNING PREFIX "_CRS %d not found" " in _PRS\n", link->irq.active); - link->irq.active = 0; +// link->irq.active = 0; } /*
GREAT! Now system boots like a charm!!! Thanks!
But using this patch you only maked acpi=noirq as default...
No, acpi=noirq entirely forbid ACPI do IRQ routing. The patch just disable a check (_CRS out of _PRS). We have no the check in previous kernel, and add it latest. Looks like a regression. In failed situation, can your VGA and USB devices work? Thanks.
Hmmm when I do init=/bin/sh and then modprobe proc usb fs, I can list devices using lspci. I don't know how to check vga - I use vga=normal in lilo.conf and it shows booting messages. But now I still see in dmesg: parport0: FIFO is stuck parport0: BUSY timeout (1) in compat_write_block_pio DMA write timed out which is shown in pci=noacpi, acpi=noirq and not shown in normal booted 2.6.4
Please verify that the system has the latest BIOS. The reason for the regression is this: ACPI: PCI Interrupt Link [LNKA] (IRQs 10 11) *3 ACPI: PCI Interrupt Link [LNKH] (IRQs 10 11) *7 ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 11 (was IRQ 3 in 2.6.4) ACPI: PCI Interrupt Link [LNKH] enabled at IRQ 11 (was IRQ 7 in 2.6.4) We used to use illegal IRQ 3 and 7, but now we use legal IRQ11 for these links. The acpi=noirq parport0 issue may be due to the bogus use of IRQ7 -- did you see that message also in 2.6.4? Just for grins, please try booting with "acpi_irq_isa=11" to see if the devices work any better on IRQ10. BTW, the DSDT has IOAPIC support, but the issue at hand is a PIC-mode failure (which should still work in any case) Is the IOAPIC enabled in the BIOS and also in the kernel build, or is there really none present? (You'll need CONFIG_X86_UP_APIC=y or build a CONFIG_SMP kernel)
Please verify that the system has the latest BIOS. --- I wrote that I have latest BIOS 1.15 The reason for the regression is this: ACPI: PCI Interrupt Link [LNKA] (IRQs 10 11) *3 ACPI: PCI Interrupt Link [LNKH] (IRQs 10 11) *7 ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 11 (was IRQ 3 in 2.6.4) ACPI: PCI Interrupt Link [LNKH] enabled at IRQ 11 (was IRQ 7 in 2.6.4) We used to use illegal IRQ 3 and 7, but now we use legal IRQ11 for these links. The acpi=noirq parport0 issue may be due to the bogus use of IRQ7 -- did you see that message also in 2.6.4? --- no, under 2.6.4 everything works like a charm Just for grins, please try booting with "acpi_irq_isa=11" to see if the devices work any better on IRQ10. --- boot with David's "patch" or without it? BTW, the DSDT has IOAPIC support, but the issue at hand is a PIC-mode failure (which should still work in any case) Is the IOAPIC enabled in the BIOS and also in the kernel build, or is there really none present? (You'll need CONFIG_X86_UP_APIC=y or build a CONFIG_SMP kernel) --- I can't user IOAPIC - if so, my kernel freezes imeditially after loading vmlinuz - noway and Acer replyed me not to fix it in their BIOS - but I reported it 6 mounths ago to add this BIOS into IOAPIC blacklist
Created attachment 3216 [details] dmesg_ioapic_2.6.7
Hmm I tried compile kernel with IOAPIC and it looks that you made some fixes, coz it boots, but as you can see, there are BIOS bugs. 2.4.26 still freezes using IOAPIC.
without David's patch kernel freezes using both acpi_irq_isa=11 and acpi_irq_isa=10 when starting pcmcia-cs services
Re: acpi_irq_isa=N Yes, you guessed correctly that this was to be run w/o David's patch. I assume that you tested acpi_irq_isa=10 and acpi_irq_isa=11 individually, for if given together they'd be a NOP. Re: CONFIG_SMP kernel So in the past when you added IOAPIC support to the kernel it would freeze after loading vmlinuz. 2.4.26 still freezes like that. But 2.6.7+IOAPIC boots and later runs into the pcmcia hang like the IOAPIC=n kernel? Must be MPS error checking (below). ACPI did not find a MADT, so according to ACPI there is no IOAPIC. (conceivably, there is a BIOS SETUP option to enable it, because there IS IOAPIC support in the DSDT) MPS takes a swing at it, but gives up: found SMP MP-table at 000f63f0 Intel MultiProcessor Specification v1.4 Virtual Wire compatibility mode. SMP mptable: bad signature [0x0]! BIOS bug, MP table errors detected!... ... disabling SMP support. (tell your hw vendor) this isn't exactly reassuring. and then No local APIC present or hardware disabled RE: IOAPIC black list I'm not aware of such a black list, can you point me to the thread? I don't see this system in the dmi_scan Local-APIC black list. (local_apic_kills_bios) So I don't know why the Local APIC is not found & enabled. can you attach your /proc/cpuinfo? 0000:02:04.0 CardBus bridge: Texas Instruments PCI1520 PC card Cardbus Controller (rev 01) pinA 0000:02:04.1 CardBus bridge: Texas Instruments PCI1520 PC card Cardbus Controller (rev 01) pinB use these links: Package (0x04) { 0x0004FFFF, 0x00, \_SB.PCI0.LPCB.LNKF, 0x00 }, Package (0x04) { 0x0004FFFF, 0x01, \_SB.PCI0.LPCB.LNKG, 0x00 }, ACPI: PCI Interrupt Link [LNKF] enabled at IRQ 10 ACPI: PCI Interrupt Link [LNKG] enabled at IRQ 10 So the issue isn't with the IRQ for the cardbus controllers themselves, they're still at IRQ 10, just like they were in 2.6.4. The change was to LNKA and LNKH: ACPI: PCI Interrupt Link [LNKA] (IRQs 10 11) *3 ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 11 Package (0x04) { 0x0002FFFF, 0x00, \_SB.PCI0.LPCB.LNKA, 0x00 }, Package (0x04) { 0x001DFFFF, 0x00, \_SB.PCI0.LPCB.LNKA, 0x00 }, 0000:00:02.0 VGA compatible controller: Intel Corp. 82852/855GM Integrated Graphics Device (rev 02) (prog-if 00 [VGA]) pinA Though it is unlikely that there is a driver loaded for this graphics device and unlikely that it actually provokes interrupts. Also, in the failure case these are put on IRQ11, which should not interfere with the Cardbus on IRQ10. ACPI: PCI Interrupt Link [LNKH] (IRQs 10 11) *7 ACPI: PCI Interrupt Link [LNKH] enabled at IRQ 11 Package (0x04) { 0x001DFFFF, 0x03, \_SB.PCI0.LPCB.LNKH, 0x00 }, Device 1d pinD is EHCI: 0000:00:1d.7 USB Controller: Intel Corp. 82801DB (ICH4) USB2 EHCI Controller (rev 03) (prog-if 20 [EHCI]) Again, even if moved from IRQ7 to IRQ11, it should not interfere with cardbus. Can you clarify which kernel version works and which fails? The CRS outside PRS change that David's patch undid was in 2.6.7, but not in 2.6.6. So if 2.6.6 failed, then the cause is elewhere. Please boot the latest working kernel with "acpi_irq_balance". It should move the IRQs like they are moved in 2.6.7. Attach the dmesg if you can. It will be interesting if it fails the same way as 2.6.7. If it works then it suggests that the root cause is elsewhere. Also, it wouldn't be a bad idea to see if 2.4.27-latest works as well as 2.4.26 for it contains the same IRQ changes that 2.6.7 does, but doesn't contain a bunch of other changes.
Re: acpi_irq_isa=N Yes, you guessed correctly that this was to be run w/o David's patch. ^^ OK I assume that you tested acpi_irq_isa=10 and acpi_irq_isa=11 individually, for if given together they'd be a NOP. ^^^ I tried acpi_irq_isa=10 and kernel freezes acpi_irq_isa=11 and kernel freezes too Re: CONFIG_SMP kernel So in the past when you added IOAPIC support to the kernel it would freeze after loading vmlinuz. 2.4.26 still freezes like that. ^^^^ Yes system wrote loading linux..... then only black screen and total freeze But 2.6.7+IOAPIC boots and later runs into the pcmcia hang like the IOAPIC=n kernel? Must be MPS error checking (below). ^^^ Yes in 2.6.7 kernel boots, but freezes when starting pcmcia-cs services (or maybe parport, coz it is loaded after pcmcia-cs) ACPI did not find a MADT, so according to ACPI there is no IOAPIC. (conceivably, there is a BIOS SETUP option to enable it, because there IS IOAPIC support in the DSDT) ^^^ my bios not allow me to change anything except boot device priority :D MPS takes a swing at it, but gives up: found SMP MP-table at 000f63f0 Intel MultiProcessor Specification v1.4 Virtual Wire compatibility mode. SMP mptable: bad signature [0x0]! BIOS bug, MP table errors detected!... ... disabling SMP support. (tell your hw vendor) this isn't exactly reassuring. and then No local APIC present or hardware disabled RE: IOAPIC black list I'm not aware of such a black list, can you point me to the thread? ^^^ will find later - but my request was probably lost?? I don't see this system in the dmi_scan Local-APIC black list. (local_apic_kills_bios) So I don't know why the Local APIC is not found & enabled. can you attach your /proc/cpuinfo? ^^^^ processor : 0 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Celeron(R) CPU 2.40GHz stepping : 9 cpu MHz : 2398.020 cache size : 128 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe cid bogomips : 4784.12 with activated cpufreq 0000:02:04.0 CardBus bridge: Texas Instruments PCI1520 PC card Cardbus Controller (rev 01) pinA 0000:02:04.1 CardBus bridge: Texas Instruments PCI1520 PC card Cardbus Controller (rev 01) pinB use these links: Package (0x04) { 0x0004FFFF, 0x00, \_SB.PCI0.LPCB.LNKF, 0x00 }, Package (0x04) { 0x0004FFFF, 0x01, \_SB.PCI0.LPCB.LNKG, 0x00 }, ACPI: PCI Interrupt Link [LNKF] enabled at IRQ 10 ACPI: PCI Interrupt Link [LNKG] enabled at IRQ 10 So the issue isn't with the IRQ for the cardbus controllers themselves, they're still at IRQ 10, just like they were in 2.6.4. The change was to LNKA and LNKH: ACPI: PCI Interrupt Link [LNKA] (IRQs 10 11) *3 ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 11 Package (0x04) { 0x0002FFFF, 0x00, \_SB.PCI0.LPCB.LNKA, 0x00 }, Package (0x04) { 0x001DFFFF, 0x00, \_SB.PCI0.LPCB.LNKA, 0x00 }, 0000:00:02.0 VGA compatible controller: Intel Corp. 82852/855GM Integrated Graphics Device (rev 02) (prog-if 00 [VGA]) pinA Though it is unlikely that there is a driver loaded for this graphics device and unlikely that it actually provokes interrupts. Also, in the failure case these are put on IRQ11, which should not interfere with the Cardbus on IRQ10. ACPI: PCI Interrupt Link [LNKH] (IRQs 10 11) *7 ACPI: PCI Interrupt Link [LNKH] enabled at IRQ 11 Package (0x04) { 0x001DFFFF, 0x03, \_SB.PCI0.LPCB.LNKH, 0x00 }, Device 1d pinD is EHCI: 0000:00:1d.7 USB Controller: Intel Corp. 82801DB (ICH4) USB2 EHCI Controller (rev 03) (prog-if 20 [EHCI]) Again, even if moved from IRQ7 to IRQ11, it should not interfere with cardbus. Can you clarify which kernel version works and which fails? ^^^^ 2.6.4, 2.4.26 works, when localapic disabled The CRS outside PRS change that David's patch undid was in 2.6.7, but not in 2.6.6. So if 2.6.6 failed, then the cause is elewhere. Please boot the latest working kernel with "acpi_irq_balance". ^^^^ I'll try It should move the IRQs like they are moved in 2.6.7. Attach the dmesg if you can. It will be interesting if it fails the same way as 2.6.7. If it works then it suggests that the root cause is elsewhere. Also, it wouldn't be a bad idea to see if 2.4.27-latest works as well as 2.4.26 for it contains the same IRQ changes that 2.6.7 does, but doesn't contain a bunch of other changes.
You were right Len, 2.6.4 acpi_irq_balance caused total freeze at the same point as 2.6.7
2.4.27-rc1 freezes as soon as 2.4.26 when localAPIC enabled: Uncompressing ....booting the kernel OK freeze
My post to linux-kernel@vger.kernel.org about ioapic can be found here: http://www.uwsg.iu.edu/hypermail/linux/kernel/0311.0/0431.html Some issues with Acer TravelMate 242 From: Michal Semler (volny.cz) Date: Tue Nov 04 2003 - 14:21:21 EST
is it possible to boot windows on this laptop and observe what IRQ assignments it uses?
Created attachment 3242 [details] HW_WinXP_Czech.txt
Sending info, but Windows installed on my laptop are in czech language. Hope this will be helpfull too
the attachment in comment #48 is a binary file. it has the signature of an MP3, but I couldn't get any sound out of it. Is it possible to paste the text instead? Basically we want to know the mapping between IRQ#'s and device names.
Hi Len, it is output of System info tool in Windows XP. You can try open it in this tool, or open it in vim. In vim, there are some strange chars between normal chars, but text is reasonable. PLS try and send me comment.
Try use Unicode and windows1250 and you'll see normal text document. Problem may be names of some devices, which are in czech language.
Created attachment 3287 [details] ANSI version of Unicode HW_WinXP_Czech.txt
IRQ 0 Syst
I'll try translate some things: IRQ 0 System timer OK IRQ 1 Standard keyboard OK IRQ 4 Fast IrDA port OK IRQ 6 Floppy disk controller OK IRQ 7 IEEE 802.11 Wireless LAN/PC Card(5V) OK IRQ 8 Real time clock OK IRQ 9 Syst
any luck with 2.6.9?
no, as you can see in kernel version :( M.
Please re-test with more recent kernel. If this is still the problem, please reopen.