Hi, On 04/03/20 19:19, Michael Christie wrote:
On 04/03/2020 09:32 AM, Bodo Stroesser wrote:/* * dev_attrib attributes for devices using the target core SBC/SPC diff --git a/drivers/target/target_core_pscsi.c b/drivers/target/target_core_pscsi.c index 4e37fa9b409d..e7d92ef43ca4 100644 --- a/drivers/target/target_core_pscsi.c +++ b/drivers/target/target_core_pscsi.c @@ -1073,6 +1073,7 @@ static const struct target_backend_ops pscsi_ops = { .transport_flags_default = TRANSPORT_FLAG_PASSTHROUGH | TRANSPORT_FLAG_PASSTHROUGH_ALUA | TRANSPORT_FLAG_PASSTHROUGH_PGR, + .transport_flags_changeable = TRANSPORT_FLAG_PASSTHROUGH_PGR, .attach_hba = pscsi_attach_hba, .detach_hba = pscsi_detach_hba, .pmode_enable_hba = pscsi_pmode_enable_hba,Do we need more code to support this? For example, if LIO core is now handling PRs then it uses the I_T nexus info from LIO core for registrations if its not provided in the PR command. But port/target INQUIRY info would be from the struct scsi_device that pscsi is using since we pass INQUIRY down to that device. We will end up with mismatches where a PR-in READ_FULL_STATUS would return different results than the INQUIRY.
You are right. There is something missing. Would it be a good idea to check and change INQUIRY response and maybe responses from further commands (MODE SENSE, ...) according to in core data? I think, such a "post processing" function for responses would be useful for tcmu also when emulating pr in core. Otherwise userspace needs to gather information from sys-FS and maybe additionally needs to have per command nexus info to provide consistent responses. To me that doesn't sound like a complete "in core pr emulation". Btw, I think it would be helpful as a first part for ALUA also. Thank you, Bodo
