On Thu, 2018-02-22 at 02:23 +0000, Damien Le Moal wrote:
On Wed, 2018-02-21 at 09:23 -0800, Bart Van Assche wrote:
[ ... ]
This does not compile.
This patch depends on another patch that is not yet in Martin's tree. See also https://marc.info/?l=linux-scsi&m=151675130615597. I should have mentioned this in the patch description.
Testing this, the rcu hang is now gone.
Thanks for the testing :-)
However, the behavior of the error recovery is still different from what I see in 4.15 and 4.14. For my test case, an unaligned write to a sequential zone on a ZAC drive connected to an AHCI port, the report zone issued during the disk revalidation after the write error fails with a timeout, which causes capacity change to 0, port reset and recovery again. Eventually, everything comes back up OK, but it takes some time.
I am investigating to make sure I am not hitting a device FW bug to confirm if this is a kernel problem.
This patch was tested with the SRP protocol. I'm not an ATA expert but I hope that someone who is more familiar with ATA than I can chime in.
Bart.