|
From: Ken M. <li...@ma...> - 2004-11-05 04:05:01
|
Anton Altaparmakov wrote: > Hi, > > [CC:-ing ntfs-dev ML since this is really a developer issue.] > > On Thu, 2004-11-04 at 07:10, Ken MacFerrin wrote: > >>I have a PC setup to dual boot Gentoo Linux & WinXP Pro SP2 and continue >>to have a segmentation fault when trying to mount one of my ntfs >>partitions under linux. I've tried this with several 2.6.8 and 2.6.9 >>kernels (currently 2.6.9-gentoo-r1) and have had the same issue each >>time. I've also tried this with the ntfs kernel driver complied both >>monolithic and as a module (current). >> >>The issue is that I can mount one of my ntfs partitions without issue >>but when trying to mount the XP OS partition (/dev/hda1) I receive a >>"Segmentation Fault" error and the mount command exits. If I try to >>mount it a second time then the command line just hangs. >> >>The odd part is that Windows boots just fine off this partition and I've >>defragmented the drive and run chkdsk (with the 'fix errors' option) >>multiple times under Win with no issues. >> >>I've provided some snipped dmesg output below while trying to mount at >>boot with the following options in fstab: >>/dev/hda1 /mnt/winxp ntfs auto,ro,owner,user,nosuid,nodev,umask=0227 0 0 >> >>Since the partition works fine natively and I can cleanly mount my other >>ntfs partition via linux I can't tell if there's genuinely an issue with >>the partition or if this is a buggy error from the driver. Any suggestions? > > > Yes! Thank you for the excellent bug report and error output. It is > nice to get such a good bug report! > > There are two bugs. One that causes it to crash. (That one is just > stupid error handling.) And one that causes it to think there is > something corrupt and aborting the read even though the corruption is in > mft record 0x12 while it is actually reading 0x10. This is really > silly. Anyway, if you apply the attached patch, it will no longer > crash. Could you do it and verify that it now does not crash? Thanks a > lot! > > To apply the patch just: cd /usr/src/linux [or wherever the kernel is] > and: patch -p1 < /path/to/ntfs-attr-fix1.diff > > After you have done that could you run chkdsk /f on that partition from > windows and then try again mounting? Assuming it still fails it would > be really useful to have a copy of the beginning of your $MFT system > file so I can replicate the problem here. For this could you install > the ntfsprogs package (if not available in gentoo, get it from > http://linux-ntfs.sourceforge.net/downloads.html) and run: > > ntfscat -fi 0 /dev/hda1 > /tmp/mymft.dat > (You can then compress the file with bzip2 -9 /tmpmymft.dat) > > If you could send me that or put it somewhere for me to download it > would be great. If the file is too big for that than copy out the > beginning like so: > > dd if=/tmp/mymft.dat of=/tmp/mymftbegin.dat bs=1024 count=32 > (Again can compress with bzip2 -9 /tmp/mymftbegin.dat) > [snip] Anton, Many thanks for your quick reply. I have applied the patch, recompiled and then ran a chkdsk (under Win) on the /dev/hda1 partition.. The good news is that when trying to mount under linux that mount no longer segfaults or hangs. The bad news is that I am still unable to mount the partition and get the errors listed below. Again, this same partition still works fine natively. I'll copy you offline with a link to my MFT. Again, any suggestions appreciated. Thanks, Ken ------------------------ Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): ntfs_end_buffer_async_read(): post_read_mst_fixup() failed, corrupt mft record 0x12. Run chkdsk. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): map_mft_record_page(): Failed with error code 5. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): map_mft_record(): Failed with error code 5. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): map_extent_mft_record(): Failed to map extent mft record, error code 5. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): ntfs_external_attr_find(): Failed to map extent mft record 0x10 of base inode 0x0. Unmount and run chkdsk. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): ntfs_read_inode_mount(): Failed to lookup $MFT/$DATA attribute extent. $MFT is corrupt. Run chkdsk. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): ntfs_read_inode_mount(): Failed. Marking inode as bad. Nov 4 20:40:35 [kernel] NTFS-fs error (device hda1): ntfs_fill_super(): Failed to load essential metadata. Nov 4 20:41:31 [kernel] NTFS-fs error (device hda1): ntfs_end_buffer_async_read(): post_read_mst_fixup() failed, corrupt mft record 0x12. Run chkdsk. Nov 4 20:41:31 [kernel] NTFS-fs error (device hda1): map_mft_record_page(): Failed with error code 5. -------------------- |
