LVM2 on 2.4.22. (Fedora core 1 with nptl)

Please post notes back to here if you managed to work out the patches that are needed. I'd like to stick with the fedora kernel instead of grabbing a clean 2.4.22 as I need some of the other patching that Fedora have done.

function)

The BLKGETSIZE64 error is due to it not being in the "userspace" <linux/fs.h> for Fedora (glibc-kernel-headers package).
Just add the #define line from /usr/src/linux/include/linux/fs.h (or compile against that fs.h as other parts of LVM2 seem to want to compile against kernel private headers from the warnings thrown during compile).

I've got a ~56K bz2 diff from after getting LVM2 working against the fedora 2.4.22-1.2115.nptl kernel sources (after make mrproper, which I always have to do to get redhat kernel's to compile).

Here the definition of working is compiles + system rebooted with kernel (LVM2 devmapper is a module). Module loaded and used pvcreate, vgextend, pvmove, on test data and all *seem* ok.

Can I mail a 56K attachment to the list, or is there somewhere else I can put it?

Thanks Lucca and others for responding to my previous request for help.

I am now trying to apply the kernel patches to Fedora Core 1 (2.4.22- 1.2115.nptl). I am using the latest version, downloaded from the sistina.com web site (device-mapper-1.00.07 and LVM2.2.00.08). I have massaged the ioctl patches, as suggested, and they have applied successfully. There were a few mismatches, but they were obvious and easily resolved.

Doing a patch --dry-run -Np1... of thelinux-2.4.22- VFS-lock patch I get this:
patching file drivers/md/lvm.c
Reversed (or previously applied) patch detected! Skipping patch. 3 out of 3 hunks ignored -- saving rejects to file drivers/md/lvm.c.rej

However, when I look at the source code, this patch is NOT applied - at least the way I understand it. I have included both the patch and the relevant snippets from the source code below. I would appreciate it if someone can confirm for me that my understanding is correct and that the patch program is lying to me. In applying the ioctl patches when it told me some of the patches were already applied I checked and they were, so I don't understand why it would get it wrong now! If my understanding is correct and the patch has not been applied, I will make the changes to the source manually. If anyone has any better ideas or explanations I would appreciate hearing them.

Thanks in advance for your help.

P.S. When (if) I get all the patches in, a compiled and bootable kernel and LVM seems to be working, I am happy to do a diff between the two kernel source trees to create patches for Fedora Core 1, if that would be of interest. Never created patches before, so I may need some help. Also, any use of the patches would be at your own risk, etc.

The patch in the linux.2.4.22-VFS-lock.patch file is:
===== drivers/md/lvm.c 1.19 vs edited =====

@@ -229,9 +229,6 @@
 #define DEVICE_OFF(device)
 #define LOCAL_END_REQUEST
 
-/* lvm_do_lv_create calls fsync_dev_lockfs()/unlockfs() */
-/* #define	LVM_VFS_ENHANCEMENT */
-
 #include <linux/config.h>
 #include <linux/module.h>
 #include <linux/kernel.h>

@@ -2171,12 +2168,8 @@
 	if (lv_ptr->lv_access & LV_SNAPSHOT) {
 		lv_t *org = lv_ptr->lv_snapshot_org, *last;
 
-		/* sync the original logical volume */
-		fsync_dev(org->lv_dev);
-#ifdef	LVM_VFS_ENHANCEMENT
 		/* VFS function call to sync and lock the filesystem */
 		fsync_dev_lockfs(org->lv_dev);
-#endif
 
 		down_write(&org->lv_lock);
 		org->lv_access |= LV_SNAPSHOT_ORG;

@@ -2201,11 +2194,9 @@
 	else
 		set_device_ro(lv_ptr->lv_dev, 1);
 
-#ifdef	LVM_VFS_ENHANCEMENT
 /* VFS function call to unlock the filesystem */
 	if (lv_ptr->lv_access & LV_SNAPSHOT)
 		unlockfs(lv_ptr->lv_snapshot_org->lv_dev);
-#endif
 
 	lvm_gendisk.part[MINOR(lv_ptr->lv_dev)].de =
 	    lvm_fs_create_lv(vg_ptr, lv_ptr);

The relevant snipets from the source file are:
Hunk 1:
#define DEVICE_OFF(device)
#define LOCAL_END_REQUEST

/* lvm_do_lv_create calls fsync_dev_lockfs()/unlockfs() */
#define LVM_VFS_ENHANCEMENT

#include <linux/config.h>
#include <linux/module.h>
#include <linux/kernel.h>

Hunk 2:

	if (lv_ptr->lv_access & LV_SNAPSHOT) {
		lv_t *org = lv_ptr->lv_snapshot_org, *last;

		/* sync the original logical volume */
		fsync_dev(org->lv_dev);
#ifdef	LVM_VFS_ENHANCEMENT
		/* VFS function call to sync and lock the filesystem */
		fsync_dev_lockfs(org->lv_dev);

#endif

		down_write(&org->lv_lock);
		org->lv_access |= LV_SNAPSHOT_ORG;

Hunk 3:

	else
		set_device_ro(lv_ptr->lv_dev, 1);

#ifdef LVM_VFS_ENHANCEMENT
/* VFS function call to unlock the filesystem */

	if (lv_ptr->lv_access & LV_SNAPSHOT)
		unlockfs(lv_ptr->lv_snapshot_org->lv_dev);

#endif

	lvm_gendisk.part[MINOR(lv_ptr->lv_dev)].de =
	    lvm_fs_create_lv(vg_ptr, lv_ptr);