Dan Williams wrote:
From: Dan Williams <[EMAIL PROTECTED]>
Enable handle_stripe5 to pass off write operations to
raid5_do_soft_blocks_ops (which can be run as a workqueue). The operations
moved are reconstruct-writes and read-modify-writes formerly handled by
compute_parity5.
Changelog:
* moved raid5_do_soft_block_ops changes into a separate patch
* changed handle_write_operations5 to only initiate write operations, which
prevents new writes from being requested while the current one is in flight
* all blocks undergoing a write are now marked locked and !uptodate at the
beginning of the write operation
* blocks undergoing a read-modify-write need a request flag to distinguish
them from blocks that are locked for reading. Reconstruct-writes still use
the R5_LOCKED bit to select blocks for the operation
* integrated the work queue Kconfig option
Signed-off-by: Dan Williams <[EMAIL PROTECTED]>
---
drivers/md/Kconfig | 21 +++++
drivers/md/raid5.c | 192 ++++++++++++++++++++++++++++++++++++++------
include/linux/raid/raid5.h | 3 +
3 files changed, 190 insertions(+), 26 deletions(-)
diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
index bf869ed..2a16b3b 100644
--- a/drivers/md/Kconfig
+++ b/drivers/md/Kconfig
@@ -162,6 +162,27 @@ config MD_RAID5_RESHAPE
There should be enough spares already present to make the new
array workable.
+config MD_RAID456_WORKQUEUE
+ depends on MD_RAID456
+ bool "Offload raid work to a workqueue from raid5d"
+ ---help---
+ This option enables raid work (block copy and xor operations)
+ to run in a workqueue. If your platform has a high context
+ switch penalty say N. If you are using hardware offload or
+ are running on an SMP platform say Y.
+
+ If unsure say, Y.
+
+config MD_RAID456_WORKQUEUE_MULTITHREAD
+ depends on MD_RAID456_WORKQUEUE && SMP
+ bool "Enable multi-threaded raid processing"
+ default y
+ ---help---
+ This option controls whether the raid workqueue will be multi-
+ threaded or single threaded.
+
+ If unsure say, Y.
In the final patch that gets merged, these configuration options should
go away. We are very anti-#ifdef in Linux, for a variety of reasons.
In this particular instance, code complexity increases and
maintainability decreases as the #ifdef forest grows.
Jeff
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
