On Tue, Jul 10, 2018 at 10:58 PM, Sreekanth Reddy <sreekanth.re...@broadcom.com> wrote: > On Mon, Jul 9, 2018 at 5:14 PM, Sreekanth Reddy > <sreekanth.re...@broadcom.com> wrote: >> On Thu, Jun 28, 2018 at 10:56 AM, Sreekanth Reddy >> <sreekanth.re...@broadcom.com> wrote: >>> On Thu, Jun 28, 2018 at 3:48 AM, Bart Van Assche <bart.vanass...@wdc.com> >>> wrote: >>>> On 06/24/18 23:10, Sreekanth Reddy wrote: >>>>> >>>>> Before calling scsi_internal_device_block_nowait() API; driver sets >>>>> sas_device_priv_data->block flag as one. And in the scsih_qcmd() >>>>> driver checks for this flag as shown below and return the commands >>>>> with host busy status. >>>>> >>>>> } else if (sas_target_priv_data->tm_busy || >>>>> sas_device_priv_data->block) >>>>> return SCSI_MLQUEUE_DEVICE_BUSY; >>>> >>>> >>>> That's exactly the kind of construct that should occur in the SCSI core or >>>> block layer core and not in a SCSI LLD. Additionally, as explained before, >>>> the construct you described above is racy. >>> >>> Can you please elaborate more in details about the racy condition >>> which you think? >>> Also if at all is their any racy condition here then we are ready to >>> work on it separately, >>> So please consider the fix which we have posted. >>> >>> Thanks, >>> Sreekanth >>> >>> >>> >>> >>>> >>>> Bart. >> >> >> Any update regarding this patch? > > Pinging once again! please consider this patch fix. this is a very > simple fix for the issue reported and doesn't have any side effort.
Any update on acceptance of this patch. This is a very simple fix for the current issue which is described in the patch header, please consider this patch. We will also review complete driver code path once again and we are ready to fix if any racy conditions found while reviewing in coming days. Thanks, Sreekanth
