On Mon, 26 Mar 2018, Mike Snitzer wrote: > On Mon, Mar 26 2018 at 2:29pm -0400, > mpatocka@xxxxxxxxxx <mpatocka@xxxxxxxxxx> wrote: > > > Hi > > > > Here I'm sending the second version of dm-bufio patches for arbitrary > > sector size. > > > > I dropped the patch that removes alloc_pages because it turns out that > > alloc_pages is faster than the slab allocator. > > I had already staged the previous set, so I need to go back over that > and update accordingly. > > > I also added a patch that removes struct bio from struct dm_buffer > > (reducing the size of dm_buffer to 1/3) and allocates the bio only when > > doing read or write. > > OK, I'll go over your new set and try to tease out any differences. > > Mike One more patch that I forgot to include. From: Mikulas Patocka <mpatocka@xxxxxxxxxx> Subject: [PATCH] dm-bufio: delete outdated comment This comment was true when dm-bufio was written, but since the kernel 4.3, bios can have arbitrary size and the driver splits them. Signed-off-by: Mikulas Patocka <mpatocka@xxxxxxxxxx> --- drivers/md/dm-bufio.c | 4 ---- 1 file changed, 4 deletions(-) Index: linux-2.6/drivers/md/dm-bufio.c =================================================================== --- linux-2.6.orig/drivers/md/dm-bufio.c 2018-03-13 16:38:32.257954000 +0100 +++ linux-2.6/drivers/md/dm-bufio.c 2018-03-13 16:38:32.247954000 +0100 @@ -540,10 +540,6 @@ static void __relink_lru(struct dm_buffe * * the memory must be direct-mapped, not vmalloced; * - * the I/O driver can reject requests spuriously if it thinks that - * the requests are too big for the device or if they cross a - * controller-defined memory boundary. - * * If the buffer is small enough (up to DM_BUFIO_INLINE_VECS pages) and * it is not vmalloced, try using the bio interface. * -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
