On 6/20/22 18:05, Christoph Hellwig wrote: > On Mon, Jun 20, 2022 at 06:02:30PM +0900, Damien Le Moal wrote: >> So reserving a tag/req to be able to do NCQ at the cost of max qd being 31 >> works for that. We could keep max qd at 32 by creating one more "fake" tag >> and having a request for it, that is, having the fake tag visible to the >> block layer as a reserved tag, as John's series is doing, but for the >> reserved tags, we actually need to use an effective tag (qc->hw_tag) when >> issuing the commands. And for that, we can reuse the tag of one of the >> failed commands. > > Take a look at the magic flush request in blk-flush.c, which is > preallocated but borrows a tag from the request that wants a pre- or > post-flush. The logic is rather ugly, but maybe it might actually > become cleaner by generalizing it a bit. Thanks. Will check. I am also looking at scsi_unjam_host() and scsi_eh_get_sense(). These reuse a scsi command to do eh operations. So I could use that too, modulo making it work outside of eh context to keep the command flow intact. -- Damien Le Moal Western Digital Research
