Re: [PATCH] notify block layer when using temporary change to cache_type

From: James Bottomley
Date: Tue May 27 2014 - 12:18:26 EST


On Tue, 2014-05-27 at 19:39 +0800, Vaughan Cao wrote:
> This is a fix for commit:
> 39c60a0948cc06139e2fbfe084f83cb7e7deae3b sd: fix array cache flushing bug causing performance problems
> We must notify the block layer via q->flush_flags after temporary change the cache_type to write through.
> If not, SYNCHRONIZE CACHE command will still be generated.
>
> Signed-off-by: Vaughan Cao <vaughan.cao@xxxxxxxxxx>
> ---
> drivers/scsi/sd.c | 12 ++++++++++++
> 1 file changed, 12 insertions(+)
>
> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> index 6146b9d..366e48b 100644
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
> @@ -144,6 +144,7 @@ sd_store_cache_type(struct device *dev, struct device_attribute *attr,

This is a weird function name. The name in the vanilla kernel is
cache_type_store().

> struct scsi_sense_hdr sshdr;
> static const char temp[] = "temporary ";
> int len;
> + unsigned flush;
>
> if (sdp->type != TYPE_DISK)
> /* no cache control on RBC devices; theoretically they
> @@ -174,6 +175,17 @@ sd_store_cache_type(struct device *dev, struct device_attribute *attr,
> if (sdkp->cache_override) {
> sdkp->WCE = wce;
> sdkp->RCD = rcd;
> +
> + /* set flush_flags to notify the block layer */
> + flush = 0;
> + if (sdkp->WCE) {
> + flush |= REQ_FLUSH;
> + if (sdkp->DPOFUA)
> + flush |= REQ_FUA;
> + }
> +
> + blk_queue_flush(sdkp->disk->queue, flush);
> +

Is there a reason you cut and paste from sd_revalidate_disk() instead of
just calling it directly?

James



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/