On 10/27/18 4:51 AM, Hannes Reinecke wrote: > On 10/25/18 11:10 PM, Jens Axboe wrote: >> ide-disk and ide-cd tested as working just fine, ide-tape and >> ide-floppy haven't. But the latter don't require changes, so they >> should work without issue. >> >> Add helper function to insert a request from a work queue, since we >> cannot invoke the blk-mq request insertion from IRQ context. >> >> Cc: David Miller <davem@xxxxxxxxxxxxx> >> Signed-off-by: Jens Axboe <axboe@xxxxxxxxx> >> --- >> drivers/ide/ide-atapi.c | 25 ++++-- >> drivers/ide/ide-cd.c | 175 +++++++++++++++++++++------------------- >> drivers/ide/ide-disk.c | 5 +- >> drivers/ide/ide-io.c | 101 +++++++++++++---------- >> drivers/ide/ide-park.c | 4 +- >> drivers/ide/ide-pm.c | 28 ++----- >> drivers/ide/ide-probe.c | 68 +++++++++++----- >> include/linux/ide.h | 13 ++- >> 8 files changed, 240 insertions(+), 179 deletions(-) >> > There are some things here which are curious (like always setting the > queue depth to 32, and the insert_head thingie), but as this is legacy > anyway: The insert head addition is needed, the alternative would be making the blk-mq locks irq safe. And that would be a performance hit that isn't acceptable. Perhaps I should make that clearer in the commit message... The 32 hard code is a bit strange, but I doubt it matters that much. It's enough to saturate the drive for sure, and it'll be twice that number if a scheduler is attached, ensuring we have plenty of request to get the merging we want. > Reviewed-by: Hannes Reinecke <hare@xxxxxxxx> Thanks! -- Jens Axboe
