Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
Right, I've now disabled every grsecurity kernel config option, apart from the overarching "Getrewted Kernel Security" one - indicating the problem is in one of the non #ifdef parts of the patch. Could this be a problem: diff -ruN linux/fs/namei.c linux/fs/namei.c --- linux/fs/namei.cSat May 19 18:02:45 2001 +++ linux/fs/namei.cTue May 29 01:23:36 2001 @@ -1851,8 +1963,6 @@ error = vfs_rename(old_dir->d_inode, old_dentry, new_dir->d_inode, new_dentry); unlock_kernel(); - - dput(new_dentry); exit4: dput(old_dentry); exit3: Thanks -tony On 26 Jun 2001 19:49:19 -0400, Theodore Tso wrote: > On Tue, Jun 26, 2001 at 12:25:32AM +0200, Daniel Phillips wrote: > > > This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ > > > to know about every filesystem feature, it will (correctly) refuse to touch > > > such a system for now. You could "tune2fs -O ^FEATURE_C4 /dev/hdX" to > > > turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will > > > break all of the directory indexes, I believe. > > > > This is what he wants, a workaround so he can fsck. However, the above > > command (on version 1.2-WIP) just gives me: > > > >Invalid filesystem option set: ^FEATURE_C4 > > > > Maybe he should just edit the source so it doesn't set the superblock flag > > for now. > > I haven't had a chance to analyze the directory index format to see if > an-dirindexing-ignorant e2fsck could do any damage to the index. It's > probably the case as long as the filesystem isn't corrupted, simply > modifying e2fsck to ignore the compatibility flag won't hurt. But > it's certainly not something I would recommend for any kind of > production operation. > > - Ted > - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
Right, I've now disabled every grsecurity kernel config option, apart from the overarching Getrewted Kernel Security one - indicating the problem is in one of the non #ifdef parts of the patch. Could this be a problem: diff -ruN linux/fs/namei.c linux/fs/namei.c --- linux/fs/namei.cSat May 19 18:02:45 2001 +++ linux/fs/namei.cTue May 29 01:23:36 2001 @@ -1851,8 +1963,6 @@ error = vfs_rename(old_dir-d_inode, old_dentry, new_dir-d_inode, new_dentry); unlock_kernel(); - - dput(new_dentry); exit4: dput(old_dentry); exit3: Thanks -tony On 26 Jun 2001 19:49:19 -0400, Theodore Tso wrote: On Tue, Jun 26, 2001 at 12:25:32AM +0200, Daniel Phillips wrote: This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ to know about every filesystem feature, it will (correctly) refuse to touch such a system for now. You could tune2fs -O ^FEATURE_C4 /dev/hdX to turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will break all of the directory indexes, I believe. This is what he wants, a workaround so he can fsck. However, the above command (on version 1.2-WIP) just gives me: Invalid filesystem option set: ^FEATURE_C4 Maybe he should just edit the source so it doesn't set the superblock flag for now. I haven't had a chance to analyze the directory index format to see if an-dirindexing-ignorant e2fsck could do any damage to the index. It's probably the case as long as the filesystem isn't corrupted, simply modifying e2fsck to ignore the compatibility flag won't hurt. But it's certainly not something I would recommend for any kind of production operation. - Ted - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Tue, Jun 26, 2001 at 12:25:32AM +0200, Daniel Phillips wrote: > > This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ > > to know about every filesystem feature, it will (correctly) refuse to touch > > such a system for now. You could "tune2fs -O ^FEATURE_C4 /dev/hdX" to > > turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will > > break all of the directory indexes, I believe. > > This is what he wants, a workaround so he can fsck. However, the above > command (on version 1.2-WIP) just gives me: > >Invalid filesystem option set: ^FEATURE_C4 > > Maybe he should just edit the source so it doesn't set the superblock flag > for now. I haven't had a chance to analyze the directory index format to see if an-dirindexing-ignorant e2fsck could do any damage to the index. It's probably the case as long as the filesystem isn't corrupted, simply modifying e2fsck to ignore the compatibility flag won't hurt. But it's certainly not something I would recommend for any kind of production operation. - Ted - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
I use debugfs to remove the flag before fsck'ing: Start debugfs. Type open -f -w /dev/ features -FEATURE_C5 -tony On 26 Jun 2001 00:25:32 +0200, Daniel Phillips wrote: > On Monday 25 June 2001 21:51, Andreas Dilger wrote: > > Daniel writes: > > > Sure, if your root partition is expendable, by all means go ahead. Ted > > > has already offered to start the required changes to e2fsck, which > > > reminds me, I have to send the promised docs. For now, just use normal > > > fsck and it will (in theory) turn the directory indexes back into normal > > > file blocks, and have no effect on inodes. > > > > This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ > > to know about every filesystem feature, it will (correctly) refuse to touch > > such a system for now. You could "tune2fs -O ^FEATURE_C4 /dev/hdX" to > > turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will > > break all of the directory indexes, I believe. > > This is what he wants, a workaround so he can fsck. However, the above > command (on version 1.2-WIP) just gives me: > >Invalid filesystem option set: ^FEATURE_C4 > > Maybe he should just edit the source so it doesn't set the superblock flag > for now. > > BTW, there doesn't seem to be a --version command in tune2fs. > - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
I use debugfs to remove the flag before fsck'ing: Start debugfs. Type open -f -w /dev/part features -FEATURE_C5 -tony On 26 Jun 2001 00:25:32 +0200, Daniel Phillips wrote: On Monday 25 June 2001 21:51, Andreas Dilger wrote: Daniel writes: Sure, if your root partition is expendable, by all means go ahead. Ted has already offered to start the required changes to e2fsck, which reminds me, I have to send the promised docs. For now, just use normal fsck and it will (in theory) turn the directory indexes back into normal file blocks, and have no effect on inodes. This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ to know about every filesystem feature, it will (correctly) refuse to touch such a system for now. You could tune2fs -O ^FEATURE_C4 /dev/hdX to turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will break all of the directory indexes, I believe. This is what he wants, a workaround so he can fsck. However, the above command (on version 1.2-WIP) just gives me: Invalid filesystem option set: ^FEATURE_C4 Maybe he should just edit the source so it doesn't set the superblock flag for now. BTW, there doesn't seem to be a --version command in tune2fs. - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Tue, Jun 26, 2001 at 12:25:32AM +0200, Daniel Phillips wrote: This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ to know about every filesystem feature, it will (correctly) refuse to touch such a system for now. You could tune2fs -O ^FEATURE_C4 /dev/hdX to turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will break all of the directory indexes, I believe. This is what he wants, a workaround so he can fsck. However, the above command (on version 1.2-WIP) just gives me: Invalid filesystem option set: ^FEATURE_C4 Maybe he should just edit the source so it doesn't set the superblock flag for now. I haven't had a chance to analyze the directory index format to see if an-dirindexing-ignorant e2fsck could do any damage to the index. It's probably the case as long as the filesystem isn't corrupted, simply modifying e2fsck to ignore the compatibility flag won't hurt. But it's certainly not something I would recommend for any kind of production operation. - Ted - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Monday 25 June 2001 21:51, Andreas Dilger wrote: > Daniel writes: > > > > On Wednesday 20 June 2001 16:59, Tony Gale wrote: > > > > > The main problem I have with this is that e2fsck doesn't know how > > > > > to deal with it - at least I haven't found a version that will. > > > > > This makes it rather difficult to use, especially for your root fs. > > > > Sure, if your root partition is expendable, by all means go ahead. Ted > > has already offered to start the required changes to e2fsck, which > > reminds me, I have to send the promised docs. For now, just use normal > > fsck and it will (in theory) turn the directory indexes back into normal > > file blocks, and have no effect on inodes. > > This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ > to know about every filesystem feature, it will (correctly) refuse to touch > such a system for now. You could "tune2fs -O ^FEATURE_C4 /dev/hdX" to > turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will > break all of the directory indexes, I believe. This is what he wants, a workaround so he can fsck. However, the above command (on version 1.2-WIP) just gives me: Invalid filesystem option set: ^FEATURE_C4 Maybe he should just edit the source so it doesn't set the superblock flag for now. BTW, there doesn't seem to be a --version command in tune2fs. -- Daniel - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
Daniel writes: > > > On Wednesday 20 June 2001 16:59, Tony Gale wrote: > > > > The main problem I have with this is that e2fsck doesn't know how to > > > > deal with it - at least I haven't found a version that will. This makes > > > > it rather difficult to use, especially for your root fs. > > Sure, if your root partition is expendable, by all means go ahead. Ted has > already offered to start the required changes to e2fsck, which reminds me, I > have to send the promised docs. For now, just use normal fsck and it will > (in theory) turn the directory indexes back into normal file blocks, and have > no effect on inodes. This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ to know about every filesystem feature, it will (correctly) refuse to touch such a system for now. You could "tune2fs -O ^FEATURE_C4 /dev/hdX" to turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will break all of the directory indexes, I believe. Cheers, Andreas -- Andreas Dilger \ "If a man ate a pound of pasta and a pound of antipasto, \ would they cancel out, leaving him still hungry?" http://www-mddsp.enel.ucalgary.ca/People/adilger/ -- Dogbert - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
On Monday 25 June 2001 11:46, Tony Gale wrote: > After some testing, removing the grsecurity patch seems to have solved > the disappearing-free-space problem. Now just need to find out why. > > On 20 Jun 2001 18:58:43 +0200, Daniel Phillips wrote: > > On Wednesday 20 June 2001 16:59, Tony Gale wrote: > > > The main problem I have with this is that e2fsck doesn't know how to > > > deal with it - at least I haven't found a version that will. This makes > > > it rather difficult to use, especially for your root fs. > > > > Good, the file format isn't finalized, this is only recommended for use > > on a test partition. > > It was a test partition, just happended to be a root one. AFAIAA, isn't > the fact that the file format may change irrelevant. You want people to > test this stuff or not? If it's not tested on a root fs then how do you > know there won't be any problems. > > Sorry, but when someone reports a problem, and then you say, well don't > test it like that then, it is really not acceptable. Sure, if your root partition is expendable, by all means go ahead. Ted has already offered to start the required changes to e2fsck, which reminds me, I have to send the promised docs. For now, just use normal fsck and it will (in theory) turn the directory indexes back into normal file blocks, and have no effect on inodes. Lets take advantage of the opportunity to test that feature. The new improved e2fsck should be capable of re-adding the indexes, so we'll get to test that too ;-) Lets continue this privately and see what's going on with your inode leak. -- Daniel - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
After some testing, removing the grsecurity patch seems to have solved the disappearing-free-space problem. Now just need to find out why. On 20 Jun 2001 18:58:43 +0200, Daniel Phillips wrote: > On Wednesday 20 June 2001 16:59, Tony Gale wrote: > > The main problem I have with this is that e2fsck doesn't know how to > > deal with it - at least I haven't found a version that will. This makes > > it rather difficult to use, especially for your root fs. > > Good, the file format isn't finalized, this is only recommended for use on a > test partition. It was a test partition, just happended to be a root one. AFAIAA, isn't the fact that the file format may change irrelevant. You want people to test this stuff or not? If it's not tested on a root fs then how do you know there won't be any problems. Sorry, but when someone reports a problem, and then you say, well don't test it like that then, it is really not acceptable. -tony - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
After some testing, removing the grsecurity patch seems to have solved the disappearing-free-space problem. Now just need to find out why. On 20 Jun 2001 18:58:43 +0200, Daniel Phillips wrote: On Wednesday 20 June 2001 16:59, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Good, the file format isn't finalized, this is only recommended for use on a test partition. It was a test partition, just happended to be a root one. AFAIAA, isn't the fact that the file format may change irrelevant. You want people to test this stuff or not? If it's not tested on a root fs then how do you know there won't be any problems. Sorry, but when someone reports a problem, and then you say, well don't test it like that then, it is really not acceptable. -tony - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
On Monday 25 June 2001 11:46, Tony Gale wrote: After some testing, removing the grsecurity patch seems to have solved the disappearing-free-space problem. Now just need to find out why. On 20 Jun 2001 18:58:43 +0200, Daniel Phillips wrote: On Wednesday 20 June 2001 16:59, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Good, the file format isn't finalized, this is only recommended for use on a test partition. It was a test partition, just happended to be a root one. AFAIAA, isn't the fact that the file format may change irrelevant. You want people to test this stuff or not? If it's not tested on a root fs then how do you know there won't be any problems. Sorry, but when someone reports a problem, and then you say, well don't test it like that then, it is really not acceptable. Sure, if your root partition is expendable, by all means go ahead. Ted has already offered to start the required changes to e2fsck, which reminds me, I have to send the promised docs. For now, just use normal fsck and it will (in theory) turn the directory indexes back into normal file blocks, and have no effect on inodes. Lets take advantage of the opportunity to test that feature. The new improved e2fsck should be capable of re-adding the indexes, so we'll get to test that too ;-) Lets continue this privately and see what's going on with your inode leak. -- Daniel - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
Daniel writes: On Wednesday 20 June 2001 16:59, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Sure, if your root partition is expendable, by all means go ahead. Ted has already offered to start the required changes to e2fsck, which reminds me, I have to send the promised docs. For now, just use normal fsck and it will (in theory) turn the directory indexes back into normal file blocks, and have no effect on inodes. This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ to know about every filesystem feature, it will (correctly) refuse to touch such a system for now. You could tune2fs -O ^FEATURE_C4 /dev/hdX to turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will break all of the directory indexes, I believe. Cheers, Andreas -- Andreas Dilger \ If a man ate a pound of pasta and a pound of antipasto, \ would they cancel out, leaving him still hungry? http://www-mddsp.enel.ucalgary.ca/People/adilger/ -- Dogbert - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Monday 25 June 2001 21:51, Andreas Dilger wrote: Daniel writes: On Wednesday 20 June 2001 16:59, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Sure, if your root partition is expendable, by all means go ahead. Ted has already offered to start the required changes to e2fsck, which reminds me, I have to send the promised docs. For now, just use normal fsck and it will (in theory) turn the directory indexes back into normal file blocks, and have no effect on inodes. This is only true without the COMPAT_DIR_INDEX flag. Since e2fsck _needs_ to know about every filesystem feature, it will (correctly) refuse to touch such a system for now. You could tune2fs -O ^FEATURE_C4 /dev/hdX to turn of the COMPAT_DIR_INDEX flag and let e2fsck go to town. That will break all of the directory indexes, I believe. This is what he wants, a workaround so he can fsck. However, the above command (on version 1.2-WIP) just gives me: Invalid filesystem option set: ^FEATURE_C4 Maybe he should just edit the source so it doesn't set the superblock flag for now. BTW, there doesn't seem to be a --version command in tune2fs. -- Daniel - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Wednesday 20 June 2001 18:02, Theodore Tso wrote: > On Wed, Jun 20, 2001 at 03:59:58PM +0100, Tony Gale wrote: > > The main problem I have with this is that e2fsck doesn't know how to > > deal with it - at least I haven't found a version that will. This makes > > it rather difficult to use, especially for your root fs. > > Getting e2fsck to deal with directory indexing is on my todo list at > this point. > > Daniel, do you have any preliminary patches to start with, or do I > need to start from scratch? No, I haven't written any user space code for this at all, the only source is the patch itself. The debug code might be helpful - show_buckets. It's incomplete though, it only shows one level of the tree. I need to do something about that. I'll also spiff up my pseudocode and data structure documentation and forward it. -- Daniel - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
On Wednesday 20 June 2001 16:59, Tony Gale wrote: > The main problem I have with this is that e2fsck doesn't know how to > deal with it - at least I haven't found a version that will. This makes > it rather difficult to use, especially for your root fs. Good, the file format isn't finalized, this is only recommended for use on a test partition. > And, since I used it, and have since stopped using it, I have a problem > in what all my disk free space disappears over a couple of days - I have > to run fsck to recover it, were it appears as deleted inodes with zero > dtime. I can't say for sure that the dir index stuff is at fault though. It could well be, I'd appreciate any more data you have on that. > I am currently using 2.4.6-pre3 without the dir patch installed. I am > using the grsecurity patch though. > > I have just upgraded to e2fsprogs-1.21 in the hope of sorting it out. If > that fails I'll revert to a clean 2.4.6-pre kernel. Other ideas welcome. > > -tony > > On 31 May 2001 18:13:43 +0200, Daniel Phillips wrote: > > Changes: > > > > - Freshen to 2.4.5 > > - EXT2_FEATURE_COMPAT_DIR_INDEX flag finalized > > - Break up ext2_add_entry for aesthetic reasons (Al Viro) > > - Handle more than 64K directories per directory (Andreas Dilger) > > - Bug fix: new inode no longer inherits index flag (Andreas Dilger) > > - Bug fix: correct handling of error on index create (Al Viro) > > > > To-Do: > > > > - More factoring of ext2_add_entry > > - Fall back to linear search in case of corrupted index > > - Finalize hash function - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Wed, Jun 20, 2001 at 03:59:58PM +0100, Tony Gale wrote: > > The main problem I have with this is that e2fsck doesn't know how to > deal with it - at least I haven't found a version that will. This makes > it rather difficult to use, especially for your root fs. Getting e2fsck to deal with directory indexing is on my todo list at this point. Daniel, do you have any preliminary patches to start with, or do I need to start from scratch? - Ted - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. And, since I used it, and have since stopped using it, I have a problem in what all my disk free space disappears over a couple of days - I have to run fsck to recover it, were it appears as deleted inodes with zero dtime. I can't say for sure that the dir index stuff is at fault though. I am currently using 2.4.6-pre3 without the dir patch installed. I am using the grsecurity patch though. I have just upgraded to e2fsprogs-1.21 in the hope of sorting it out. If that fails I'll revert to a clean 2.4.6-pre kernel. Other ideas welcome. -tony On 31 May 2001 18:13:43 +0200, Daniel Phillips wrote: > Changes: > > - Freshen to 2.4.5 > - EXT2_FEATURE_COMPAT_DIR_INDEX flag finalized > - Break up ext2_add_entry for aesthetic reasons (Al Viro) > - Handle more than 64K directories per directory (Andreas Dilger) > - Bug fix: new inode no longer inherits index flag (Andreas Dilger) > - Bug fix: correct handling of error on index create (Al Viro) > > To-Do: > > - More factoring of ext2_add_entry > - Fall back to linear search in case of corrupted index > - Finalize hash function > - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. And, since I used it, and have since stopped using it, I have a problem in what all my disk free space disappears over a couple of days - I have to run fsck to recover it, were it appears as deleted inodes with zero dtime. I can't say for sure that the dir index stuff is at fault though. I am currently using 2.4.6-pre3 without the dir patch installed. I am using the grsecurity patch though. I have just upgraded to e2fsprogs-1.21 in the hope of sorting it out. If that fails I'll revert to a clean 2.4.6-pre kernel. Other ideas welcome. -tony On 31 May 2001 18:13:43 +0200, Daniel Phillips wrote: Changes: - Freshen to 2.4.5 - EXT2_FEATURE_COMPAT_DIR_INDEX flag finalized - Break up ext2_add_entry for aesthetic reasons (Al Viro) - Handle more than 64K directories per directory (Andreas Dilger) - Bug fix: new inode no longer inherits index flag (Andreas Dilger) - Bug fix: correct handling of error on index create (Al Viro) To-Do: - More factoring of ext2_add_entry - Fall back to linear search in case of corrupted index - Finalize hash function - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Wed, Jun 20, 2001 at 03:59:58PM +0100, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Getting e2fsck to deal with directory indexing is on my todo list at this point. Daniel, do you have any preliminary patches to start with, or do I need to start from scratch? - Ted - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [UPDATE] Directory index for ext2
On Wednesday 20 June 2001 16:59, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Good, the file format isn't finalized, this is only recommended for use on a test partition. And, since I used it, and have since stopped using it, I have a problem in what all my disk free space disappears over a couple of days - I have to run fsck to recover it, were it appears as deleted inodes with zero dtime. I can't say for sure that the dir index stuff is at fault though. It could well be, I'd appreciate any more data you have on that. I am currently using 2.4.6-pre3 without the dir patch installed. I am using the grsecurity patch though. I have just upgraded to e2fsprogs-1.21 in the hope of sorting it out. If that fails I'll revert to a clean 2.4.6-pre kernel. Other ideas welcome. -tony On 31 May 2001 18:13:43 +0200, Daniel Phillips wrote: Changes: - Freshen to 2.4.5 - EXT2_FEATURE_COMPAT_DIR_INDEX flag finalized - Break up ext2_add_entry for aesthetic reasons (Al Viro) - Handle more than 64K directories per directory (Andreas Dilger) - Bug fix: new inode no longer inherits index flag (Andreas Dilger) - Bug fix: correct handling of error on index create (Al Viro) To-Do: - More factoring of ext2_add_entry - Fall back to linear search in case of corrupted index - Finalize hash function - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Ext2-devel] Re: [UPDATE] Directory index for ext2
On Wednesday 20 June 2001 18:02, Theodore Tso wrote: On Wed, Jun 20, 2001 at 03:59:58PM +0100, Tony Gale wrote: The main problem I have with this is that e2fsck doesn't know how to deal with it - at least I haven't found a version that will. This makes it rather difficult to use, especially for your root fs. Getting e2fsck to deal with directory indexing is on my todo list at this point. Daniel, do you have any preliminary patches to start with, or do I need to start from scratch? No, I haven't written any user space code for this at all, the only source is the patch itself. The debug code might be helpful - show_buckets. It's incomplete though, it only shows one level of the tree. I need to do something about that. I'll also spiff up my pseudocode and data structure documentation and forward it. -- Daniel - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/