| Message ID | 20210823202930.137278-2-mcgrof@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | block: first batch of add_disk() error handling conversions | expand |
On Mon, Aug 23, 2021 at 01:29:21PM -0700, Luis Chamberlain wrote: > The single put_disk() is useful if you know you're not doing > a cleanup after add_disk(), but since we want to add support > for that, just use the normal form of blk_cleanup_disk() to > cleanup the queue and put the disk. Hmm, I don't think this is correct. The request_queue is owned by the core SCSI code.
On Tue, Aug 24, 2021 at 06:52:53AM +0100, Christoph Hellwig wrote: > On Mon, Aug 23, 2021 at 01:29:21PM -0700, Luis Chamberlain wrote: > > The single put_disk() is useful if you know you're not doing > > a cleanup after add_disk(), but since we want to add support > > for that, just use the normal form of blk_cleanup_disk() to > > cleanup the queue and put the disk. > > Hmm, I don't think this is correct. The request_queue is owned by the > core SCSI code. Alright, I'll drop this one. For the life of me I can't find the respective probe call on the scsi layer. Luis
On Fri, Aug 27, 2021 at 11:27:36AM -0700, Luis Chamberlain wrote: > On Tue, Aug 24, 2021 at 06:52:53AM +0100, Christoph Hellwig wrote: > > On Mon, Aug 23, 2021 at 01:29:21PM -0700, Luis Chamberlain wrote: > > > The single put_disk() is useful if you know you're not doing > > > a cleanup after add_disk(), but since we want to add support > > > for that, just use the normal form of blk_cleanup_disk() to > > > cleanup the queue and put the disk. > > > > Hmm, I don't think this is correct. The request_queue is owned by the > > core SCSI code. > > Alright, I'll drop this one. For the life of me I can't find the respective > probe call on the scsi layer. What probe call? SCSI allocates the request_queue using the normal blk_mq_init_queue function in scsi_alloc_sdev. It it then used to send SCSI passthrough commands for probing before eventually binding an upper level driver using the driver model (or something not binding one at all if there is none for the device type).
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 610ebba0d66e..7d5217905374 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3508,7 +3508,7 @@ static int sd_probe(struct device *dev) out_free_index: ida_free(&sd_index_ida, index); out_put: - put_disk(gd); + blk_cleanup_disk(gd); out_free: sd_zbc_release_disk(sdkp); kfree(sdkp);
The single put_disk() is useful if you know you're not doing a cleanup after add_disk(), but since we want to add support for that, just use the normal form of blk_cleanup_disk() to cleanup the queue and put the disk. Signed-off-by: Luis Chamberlain <mcgrof@kernel.org> --- drivers/scsi/sd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)