floppy hang with 2.4.9-ac1x
Using mtools, such as mdir and mformat, under 2.4.9-ac16 (from 2.4.9-ac10 onwards, IIRC) hard-hangs my linux system....
dual P3/700, SMP kernel, gcc-3.0.1, Abit VP6 with VIA 694X chipset, mtools-3.9.7
dmesg about the floppy:
Floppy drive(s): fd0 is 1.44M
FDC 0 is a post-1991 82077
/proc/dma:
4: cascade
/proc/interrupts:
CPU0 CPU1
0: 42356 44396 IO-APIC-edge timer 1: 2122 3557 IO-APIC-edge keyboard 2: 0 0 XT-PIC cascade 8: 0 3 IO-APIC-edge rtc 14: 6150 3406 IO-APIC-edge ide0 16: 33 32 IO-APIC-level sym53c8xx 17: 23 21 IO-APIC-level sym53c8xx, sym53c8xx, EMU10K1 18: 12680 12663 IO-APIC-level ide2, ide3, DE500-AA (eth0) 19: 0 0 IO-APIC-level usb-uhci, usb-uhci NMI: 0 0 LOC: 86651 86651 ERR: 0 MIS: 0
The floppy-drive itself functions quite well in DOS, so it's not that.
The symptoms are easy to describe:
the console hangs, and switching to another doesn't work. Also the num-lock key is dead. The floppy light stays on, but it doesn't sound like it is doing anything but spinning the motor.
There are no messages in the system logs.
Any hints on what to do?
Thanks,
Jurriaan
Signature · 5 lines
Cadets were humping in the corridors, a party of outraged Senators roamed Krane barracks, everyone, including me, questioned orders, ... Law and order. David Feintuch - Fisherman's Hope. GNU/Linux 2.4.9-ac16 SMP/ReiserFS 2x1402 bogomips load av: 0.00 0.03 0.04
Quoted 1 line
2.4.9-ac10 onwards, IIRC) hard-hangs my linux system....
Knowing exactly which release would be useful
Quoted 2 lines
dual P3/700, SMP kernel, gcc-3.0.1, Abit VP6 with VIA 694X chipset, mtools-3.9.7
Does it occur with gcc 2.95/2.96 ?
Quoted 6 lines
the console hangs, and switching to another doesn't work. Also the num-lock key is dead. The floppy light stays on, but it doesn't sound like it is doing anything but spinning the motor.
There are no messages in the system logs.Any hints on what to do?
Try an older gcc to eliminate that case (I dont think it is the guilty party but I dont trust the memory barriers in such an old ugly driver to be right) and then find out which precise release broke it
Alan
Re: (lkml)Re: floppy hang with 2.4.9-ac1x
In reply to [email protected]Quoted 3 lines · Alan Cox
Quoted 1 line
2.4.9-ac10 onwards, IIRC) hard-hangs my linux system....
Knowing exactly which release would be useful
It worked with 2.4.9-ac15 and before, and it doesn't work with 2.4.9-ac16.
Quoted 5 lines
Quoted 3 lines
dual P3/700, SMP kernel, gcc-3.0.1, Abit VP6 with VIA 694X chipset, mtools-3.9.7
Does it occur with gcc 2.95/2.96 ?
yes, 2.4.9-ac16 crashes both with gcc-3.0.1 and with gcc-2.95.2.
Quoted 6 lines
Quoted 6 lines
the console hangs, and switching to another doesn't work. Also the num-lock key is dead. The floppy light stays on, but it doesn't sound like it is doing anything but spinning the motor.
There are no messages in the system logs.
Quoted 6 lines
Quoted 1 line
Any hints on what to do?
Try an older gcc to eliminate that case (I dont think it is the guilty party but I dont trust the memory barriers in such an old ugly driver to be right) and then find out which precise release broke it
It's 2.4.9-ac16 that broke it, and I've straced it. This is in 2.4.9-ac15:
open("/home/jurriaan/.mtoolsrc", O_RDONLY) = -1 ENOENT (No such file or directory) rt_sigaction(SIGHUP, {0x805ff50, [HUP], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 rt_sigaction(SIGINT, {0x805ff50, [INT], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 rt_sigaction(SIGTERM, {0x805ff50, [TERM], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 rt_sigaction(SIGQUIT, {0x805ff50, [QUIT], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 stat64("/home/jurriaan/.mcwd", 0xbffff37c) = -1 ENOENT (No such file or directory) open("/dev/fd0", O_RDONLY|O_EXCL|O_LARGEFILE) = 4fstat64(4, {st_mode=S_IFBLK|0660, st_rdev=makedev(2, 0), ...}) = 0flock(4, LOCK_SH|LOCK_NB) = 0ioctl(4, FDGETPRM, 0xbfffddd0) = 0read(4, "\353V\220DRDOS 7\0\2\1\1\0\2\340\0@\v\360\t\0\22\0\2\0"..., 256) = 256ioctl(4, FDGETPRM, 0xbfffe5e0) = 0old_mmap(NULL, 151552, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4014d000
brk(0x8075000) = 0x8075000_llseek(4, 512, [512], SEEK_SET) = 0
read(4, "\360\377\377\3@\0\5`\0\7\200\0\t\240\0\v\300\0\r\340\0"..., 17920) = 17920open("/etc/localtime", O_RDONLY) = 5fstat64(5, {st_mode=S_IFREG|0644, st_size=1058, ...}) = 0old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x40017000 read(5, "TZif\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\v\0\0\0\v\0"..., 4096) = 1058
close(5) = 0
and in 2.4.9-ac16 I see:
open("/home/jurriaan/.mtoolsrc", O_RDONLY) = -1 ENOENT (No such file or directory) rt_sigaction(SIGHUP, {0x805ff50, [HUP], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 rt_sigaction(SIGINT, {0x805ff50, [INT], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 rt_sigaction(SIGTERM, {0x805ff50, [TERM], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 rt_sigaction(SIGQUIT, {0x805ff50, [QUIT], SA_RESTART|0x4000000}, {SIG_DFL}, 8) = 0 stat64("/home/jurriaan/.mcwd", 0xbffff37c) = -1 ENOENT (No such file or directory) open("/dev/fd0", O_RDONLY|O_EXCL|O_LARGEFILE) = 4fstat64(4, {st_mode=S_IFBLK|0660, st_rdev=makedev(2, 0), ...}) = 0flock(4, LOCK_SH|LOCK_NB) = 0ioctl(4, FDGETPRM, 0xbfffddd0) = 0read(4, "\353V\220DRDOS 7\0\2\1\1\0\2\340\0@\v\360\t\0\22\0\2\0"..., 256) = 256ioctl(4, FDGETPRM, 0xbfffe5e0) = 0old_mmap(NULL, 151552, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4014d000
brk(0x8075000) = 0x8075000_llseek(4, 512, [512], SEEK_SET) = 0
read(4,
and there it hangs. Also, in 2.4.9-ac16 it is reading from fd 3 for some reason, but I couldn't quite see why. The values in the ioctl lines etc. are the same. Exchanging mtools-3.9.7 with 3.9.8-pre6 (the latest release) didn't change anything.
Good luck,
Jurriaan
Signature · 3 lines
Whoever thought that one up should have his head examined. Bloody minstrels... Simon R Green - Blue Moon Rising GNU/Linux 2.4.9-ac16 SMP/ReiserFS 2x1402 bogomips load av: 0.57 0.15 0.05
Re: (lkml)Re: floppy hang with 2.4.9-ac1x
Quoted 4 lines
Quoted 1 line
Knowing exactly which release would be useful
It worked with 2.4.9-ac15 and before, and it doesn't work with 2.4.9-ac16.
Ok that sounds like another instance of the Ingo ksoftirq changes breaking stuff - thanks