kernel.git/drivers/md/dm-raid1.c, branch linux-2.6.23.y

some kmalloc/memset ->kzalloc (tree wide)

2007-07-19T17:04:50Z

Transform some calls to kmalloc/memset to a single kzalloc (or kcalloc). Here is a short excerpt of the semantic patch performing this transformation: @@ type T2; expression x; identifier f,fld; expression E; expression E1,E2; expression e1,e2,e3,y; statement S; @@ x = - kmalloc + kzalloc (E1,E2) ... when != \(x->fld=E;\|y=f(...,x,...);\|f(...,x,...);\|x=E;\|while(...) S\|for(e1;e2;e3) S\) - memset((T2)x,0,E1); @@ expression E1,E2,E3; @@ - kzalloc(E1 * E2,E3) + kcalloc(E1,E2,E3) [akpm@linux-foundation.org: get kcalloc args the right way around] Signed-off-by: Yoann Padioleau Cc: Richard Henderson Cc: Ivan Kokshaysky Acked-by: Russell King Cc: Bryan Wu Acked-by: Jiri Slaby Cc: Dave Airlie Acked-by: Roland Dreier Cc: Jiri Kosina Acked-by: Dmitry Torokhov Cc: Benjamin Herrenschmidt Acked-by: Mauro Carvalho Chehab Acked-by: Pierre Ossman Cc: Jeff Garzik Cc: "David S. Miller" Acked-by: Greg KH Cc: James Bottomley Cc: "Antonino A. Daplas" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

dm raid1: handle log failure

2007-07-12T22:01:08Z

When writing to a mirror, the log must be updated first. Failure to update the log could result in the log not properly reflecting the state of the mirror if the machine should crash. We change the return type of the rh_flush function to give us the ability to check if a log write was successful. If the log write was unsuccessful, we fail the writes to avoid the case where the log does not properly reflect the state of the mirror. A follow-up patch - which is dependent on the ability to requeue I/O's to core device-mapper - will requeue the I/O's for retry (allowing the mirror to be reconfigured.) Signed-off-by: Jonathan Brassow Signed-off-by: Alasdair G Kergon Signed-off-by: Linus Torvalds

dm raid1: handle resync failures

2007-07-12T22:01:08Z

Device-mapper mirroring currently takes a best effort approach to recovery - failures during mirror synchronization are completely ignored. This means that regions are marked 'in-sync' and 'clean' and removed from the hash list. Future reads and writes that query the region will incorrectly interpret the region as in-sync. This patch handles failures during the recovery process. If a failure occurs, the region is marked as 'not-in-sync' (aka RH_NOSYNC) and added to a new list 'failed_recovered_regions'. Regions on the 'failed_recovered_regions' list are not marked as 'clean' upon removal from the list. Furthermore, if the DM_RAID1_HANDLE_ERRORS flag is set, the region is marked as 'not-in-sync'. This action prevents any future read-balancing from choosing an invalid device because of the 'not-in-sync' status. If "handle_errors" is not specified when creating a mirror (leaving the DM_RAID1_HANDLE_ERRORS flag unset), failures will be ignored exactly as they would be without this patch. This is to preserve backwards compatibility with user-space tools, such as 'pvmove'. However, since future read-balancing policies will rely on the correct sync status of a region, a user must choose "handle_errors" when using read-balancing. Signed-off-by: Jonathan Brassow Signed-off-by: Alasdair G Kergon Signed-off-by: Linus Torvalds

dm raid1: clear region outside spinlock

2007-07-12T22:01:08Z

A clear_region function is permitted to block (in practice, rare) but gets called in rh_update_states() with a spinlock held. The bits being marked and cleared by the above functions are used to update the on-disk log, but are never read directly. We can perform these operations outside the spinlock since the bits are only changed within one thread viz. - mark_region in rh_inc() - clear_region in rh_update_states(). So, we grab the clean_regions list items via list_splice() within the spinlock and defer clear_region() until we iterate over the list for deletion - similar to how the recovered_regions list is already handled. We then move the flush() call down to ensure it encapsulates the changes which are done by the later calls to clear_region(). Signed-off-by: Jonathan Brassow Signed-off-by: Alasdair G Kergon Signed-off-by: Linus Torvalds

dm raid1: fix status

2007-07-12T22:01:08Z

Fix mirror status line broken in dm-log-report-fault-status.patch: - space missing between two words - placeholder ("0") required for compatibility with a subsequent patch - incorrect offset parameter Cc: stable@kernel.org Signed-off-by: Milan Broz Signed-off-by: Alasdair G Kergon Signed-off-by: Linus Torvalds

dm: remove duplicate module name from error msgs

2007-07-12T22:01:08Z

Remove explicit module name from messages as the macro now includes it automatically. Signed-off-by: Alasdair G Kergon Signed-off-by: Linus Torvalds

dm raid1: switch rh_in_sync to blocking in do_reads

2007-05-09T19:30:48Z

The call to rh_in_sync() in do_reads() should be allowed to block. It is in the mirror worker thread which already permits blocking operations. This will be needed to support clustered mirroring which will perform network operations. Signed-off-by: Jonathan Brassow Signed-off-by: Alasdair G Kergon Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

dm raid1: fix to commit pending clear region requests

2007-05-09T19:30:48Z

With the code as it is, it is possible for oustanding clear region requests never to get flushed when a mirror is deactivated or suspended. This means there will always be some resync work required when a mirror is activated, even though it may very well be in-sync. Always requesting the flush doesn't hurt us. This is because the log tracks whether any changes occurred and, if not, no flush is performed. Signed-off-by: Jonathan Brassow Signed-off-by: Alasdair G Kergon Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

dm raid1: update dm io interface

2007-05-09T19:30:47Z

This patch ports dm-raid1.c to the new dm-io interface. Signed-off-by: Milan Broz Signed-off-by: Alasdair G Kergon Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

dm raid1: add handle_errors feature flag

2007-05-09T19:30:47Z

This patch adds the ability to specify desired features in the mirror constructor/mapping table. The first feature of interest is "handle_errors". Currently, mirroring will ignore any I/O errors from the devices. Subsequent patches will check for this flag and handle the errors. If flag/feature is not present, mirror will do nothing - maintaining backwards compatibility. Signed-off-by: Jonathan E Brassow Signed-off-by: Alasdair G Kergon Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds