On 6/6/21 11:35 PM, Christoph Hellwig wrote: > On Fri, Jun 04, 2021 at 07:52:35AM -0700, Bart Van Assche wrote: >> Damien is right. bd_disk can be NULL. From > > bd_disk is initialized in bdev_alloc, so it should never be NULL. > bi_bdev OTOH is only set afer bio_add_page in various places or not at > all in case of passthrough bios. Which is a bit of a mess and I have > plans to fix it. Hi Christoph, Thank you for having shared your plans for how to improve how bi_bdev is set. In case you would not yet have had the time to do this, please take a look at the call trace available on https://lore.kernel.org/linux-block/20210425043020.30065-1-bvanassche@xxxxxxx/. That call trace shows how bio_add_pc_page() is called by the SCSI core before alloc_disk() is called. I think that sending a SCSI command before alloc_disk() is called is fundamental in the SCSI core because the SCSI INQUIRY command has to be sent before it is known whether or not a SCSI LUN represents a disk. Thanks, Bart.
