diff "lk 2.4 udf.ko" "lk 2.6 udf.ko"
Subsystems (2)
Quoted 7 lines
By now everyone here has heard ad nauseum,
summary of how to show udf.ko sometimes cannot
read back nonzero appended to two or more open
files is:http://marc.theaimsgroup.com/?l=linux-fsdevel&m=107004461717172
Subject: Re: zeroes read back more often than appended
Today I thought to swap hard drives and try a fresh Red Hat 9 install.
I see udf.ko NOT broken back near 2.4.20. Not even when appending to two or more files in round robin sequence. I'll report back when I reach 2.6 again.
Pat LaVarre
$ date; time fswhd 767 1 mkudffs 0xFF600 0xC8 0x2 Mon Dec 1 16:54:33 MST 2003 Linux 2.4.20-8smp i686 /home/pat dd if=/dev/zero of=dd.bin bs=1M seek=767 count=1 2>/dev/null -rw-rw-r-- 1 pat pat 805306368 Dec 1 16:54 dd.bin mkudffs 1.0.0b2 for UDF FS 1.0.0-cvs, 2002/02/09 0 hexdump -C whd0.bin | head -3 00000000 aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa |................| * 0c783000 hexdump -C whd1.bin | head -3 00000000 aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa |................| * 0c783000
real 0m46.848s user 0m7.540s sys 0m7.850s
$
Quoted 5 lines
Quoted 3 lines
... udf.ko ...
http://marc.theaimsgroup.com/?l=linux-fsdevel&m=107004461717172
Subject: Re: zeroes read back more often than appendedNOT broken back near 2.4.20
Substituting new HDD for old did Not change results.
udf.ko working here includes: 2.4.23, rh 2.4.20-24.9smp, rh 2.4.20-8smp
udf.ko broken here includes: 2.6.0-test11
Pat LaVarre
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
... udf.ko ...
http://marc.theaimsgroup.com/?l=linux-fsdevel&m=107004461717172
Subject: Re: zeroes read back more often than appended...
udf.ko working here includes: 2.4.23, rh 2.4.20-24.9smp, rh 2.4.20-8smp udf.ko broken here includes: 2.6.0-test11
The binary-ish search begins ...
I see kernel.org archives 2.5.0 thru 2.5.75.
I see linux-2.5.37 arrives from kernel.org ready to try writing udf:
$ make defconfig ... $ grep UDF .config CONFIG_UDF_FS=m CONFIG_UDF_RW=y $
Pat LaVarre
Quoted 8 lines
I see linux-2.5.37 arrives from kernel.org ready to try writing udf:
$ make defconfig ... $ grep UDF .config CONFIG_UDF_FS=m CONFIG_UDF_RW=y $
Ouch, sorry, this result I cannot easily repeat. Possibly make defconfig and/or make xconfig is producing inconsistent .config, depending on in which config those commands are entered.
Pat LaVarre
Quoted 10 lines
Quoted 10 lines
Quoted 3 lines
Quoted 1 line
2.6.0-test11 [fails]
...
2.4.23 [ok]...
From: http://lxr.linux.no/source/Documentation/BUG-HUNTING
...
My apologies to Linus and the other kernel hackers for describing this brute force approach, it's hardly what a kernel hacker would do. However, it does work and it lets non-hackers help fix bugs.
Quoted 2 lines
...
The binary-ish search begins ...
The window narrows.
Somewhere between 2.5.26 and 2.5.56, we changed udf.ko from ok to fail.
More specifically, captured and sorted today, my results were:
2.4.23 udf ok
2.5.1 no defconfig
2.5.15 no defconfig
2.5.26 no gui, udf ok
2.5.37 does not boot
2.5.47 does not loop and complains much of hdc
2.5.51 does not boot
2.5.56 fails
2.6.0-test11 fails
My next try may be 2.5.41, chosen by way of dc -e '26 56 + 2 / p'
Pat LaVarre
Quoted 3 lines
The window narrows.
Somewhere between 2.5.26 and 2.5.56,
we changed udf.ko from ok to fail.
Make that between 2.5.36 and 2.5.49.
Binary search says next try 42.
Quoted 2 lines
2.5.37 does not boot
2.5.47 does not loop and complains much of hdc
2.5.36 udf ok
2.5.41 does not boot ("Attempted to kill init!")
2.5.42 does not boot ("Attempted to kill init!")
2.5.49 no gui, udf miscompares
Pat LaVarre
P.S. 2.5.41 transcribed by hand:
Stack: c0...
c0...
c0...
Call Trace:
[<c0... usbfs_add_bus+...... printk+...
... usb_register_bus+...
... usb_hcd_pci_probe+...
... pci_device_probe+...
... probe+...
... found_match+...
... do_driver_attach+...
... bus_for_each_dev+...
... driverfs_create+...
... driver_attach+...
... do_driver_attach+...
... driver_register+...
... pci_register_driver+...
... init+...
... init+...
... kernel_thread_helper+...
Code: 8b 50 ... <0>Kernel panic: Attempted to kill init!
Quoted 1 line
Make that between 2.5.36 and 2.5.49.
Maybe by now we know we broke udf.ko, not by changing udf.ko, but rather by changing something in the defconfig or in the dynamic behaviour of Linux. Specifically, I see:
2.5.36 ok once, ok again.
2.5.46 fails.
2.5.46 fs/udf/ makes and runs in 2.5.36 and then is ok, except that to make in 2.5.36, 2.5.46 fs/udf/ needs the seemingly non-substantive kluge of a patch quoted below.
Tentatively I conclude: "we broke udf.ko, not by changing udf.ko, but rather by changing something in the defconfig or in the dynamic behaviour of Linux".
Pat LaVarre
P.S. 2.5.37 thru 2.5.45 not makeable or not bootable except for 2.5.38 and 2.5.40 as yet unevaluated.
fs/udf/udf_sb.h +2 -0
@@ -1,3 +1,5 @@ +#define s_fs_info u.generic_sbp +#ifndef __LINUX_UDF_SB_H#define __LINUX_UDF_SB_H