[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <e102ba93-8fb9-1754-1881-0cd322c808d6@oracle.com>
Date: Tue, 24 Oct 2023 11:07:19 +0100
From: John Garry <john.g.garry@...cle.com>
To: Hannes Reinecke <hare@...e.de>, jejb@...ux.ibm.com,
martin.petersen@...cle.com, chenxiang66@...ilicon.com,
artur.paszkiewicz@...el.com, yanaijie@...wei.com,
jinpu.wang@...ud.ionos.com
Cc: linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org,
dlemoal@...nel.org
Subject: Re: [PATCH 07/10] scsi: libsas: Delete sas_ssp_task.task_prio
On 09/10/2023 07:55, Hannes Reinecke wrote:
>> diff --git a/include/scsi/libsas.h b/include/scsi/libsas.h
>> index d77db53cbd8d..5b2e6932c564 100644
>> --- a/include/scsi/libsas.h
>> +++ b/include/scsi/libsas.h
>> @@ -565,7 +565,6 @@ enum task_attribute {
>> struct sas_ssp_task {
>> u8 LUN[8];
>> enum task_attribute task_attr;
>> - u8 task_prio;
>> struct scsi_cmnd *cmd;
>> };
> Hmm. While true in general, we lose information on how the 'prio' field
> is coded for the various drivers.
> Would it be an idea to use explicitly state TASK_ATTR_SIMPLE
> in the various drivers such that we know what to fix if we ever
> decide to bring task attributes back to life?
The HW encoding for these drivers is the same as in the SAS COMMAND
frame - Command information unit structure, right? So the amount of info
lost is reduced. And we also have source control history.
I'd be more inclined to get rid of setting task attributes similarly as
well.
Thanks,
John
Powered by blists - more mailing lists