On Fri, 2018-10-26 at 15:58 +-0200, Hannes Reinecke wrote: +AD4 On 10/2/18 11:52 PM, Bart Van Assche wrote: +AD4 +AD4 Hello Martin, +AD4 +AD4 +AD4 +AD4 During the 2018 edition of LSF/MM there was a session about increasing SCSI +AD4 +AD4 disk probing concurrency. This patch series implements what has been proposed +AD4 +AD4 during that session, namely: +AD4 +AD4 - Make sure that the driver core is aware of asynchronous SCSI LUN probing. +AD4 +AD4 - Avoid unnecessary serialization between sd+AF8-probe() and sd+AF8-remove() because +AD4 +AD4 this could lead to a deadlock. +AD4 +AD4 +AD4 +AD4 Please consider this patch series for kernel v4.20. +AD4 +AD4 +AD4 +AD4 Thanks, +AD4 +AD4 +AD4 +AD4 Bart. +AD4 +AD4 +AD4 +AD4 Bart Van Assche (3): +AD4 +AD4 sd: Rely on the driver core for asynchronous probing +AD4 +AD4 sd: Inline sd+AF8-probe+AF8-part2() +AD4 +AD4 +AF8AXw-device+AF8-release+AF8-driver(): Do not wait for asynchronous probing +AD4 +AD4 +AD4 +AD4 drivers/base/dd.c +AHw 3 -- +AD4 +AD4 drivers/scsi/scsi.c +AHw 14 ----- +AD4 +AD4 drivers/scsi/scsi+AF8-pm.c +AHw 22 +-------- +AD4 +AD4 drivers/scsi/scsi+AF8-priv.h +AHw 3 -- +AD4 +AD4 drivers/scsi/sd.c +AHw 110 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+------------------------ +AD4 +AD4 5 files changed, 46 insertions(+-), 106 deletions(-) +AD4 +AD4 +AD4 +AD4 Initially it looks ok-ish. +AD4 But I've has so many issues with asynchronous probing that I'd be really +AD4 careful here. +AD4 However, inlining sd+AF8-probe+AF8-async() should amount for some contention to +AD4 be removed+ADs but I'd rather give it some more testing here. +AD4 I've found that testing on large configurations on iSER or SRP tend to +AD4 excite the issues. +AD4 Can you check against those, too? +AD4 I'll see to give it a spin on my test bed. Hi Hannes, As you may have noticed the build bot was not very happy with the patches in v2 of this series for the device driver core. Additional testing is welcome but please start from the patches in the following branch for any tests: https://github.com/bvanassche/linux/tree/for-next. On that branch my patches for the device driver core have been replaced with patches from Alexander Duyck. Thanks, Bart.
