From: Bart Van Assche <bvanassche@acm.org>
To: "Bao D. Nguyen" <quic_nguyenb@quicinc.com>, Jens Axboe <axboe@kernel.dk>
Cc: linux-block@vger.kernel.org, linux-scsi@vger.kernel.org,
"Martin K . Petersen" <martin.petersen@oracle.com>,
Christoph Hellwig <hch@lst.de>, Can Guo <quic_cang@quicinc.com>,
Avri Altman <avri.altman@wdc.com>,
Damien Le Moal <dlemoal@kernel.org>,
Ming Lei <ming.lei@redhat.com>,
"James E.J. Bottomley" <jejb@linux.ibm.com>,
Stanley Chu <stanley.chu@mediatek.com>,
Bean Huo <beanhuo@micron.com>,
Asutosh Das <quic_asutoshd@quicinc.com>,
Arthur Simchaev <Arthur.Simchaev@wdc.com>
Subject: Re: [PATCH v9 17/17] scsi: ufs: Inform the block layer about write ordering
Date: Fri, 18 Aug 2023 10:56:49 -0700 [thread overview]
Message-ID: <06d34c2f-a89e-3134-d78f-3c071aa490cf@acm.org> (raw)
In-Reply-To: <8299025e-e895-8fdb-c62d-80f1d4c9b64c@quicinc.com>
On 8/17/23 17:19, Bao D. Nguyen wrote:
> For example, in SDB mode, after the probe and you want to enable auto-hibern8, you would call ufshcd_auto_hibern8_update() which then calls
> ufshcd_update_preserves_write_order(). Before auto-hibern8 is enabled, you would check this condition:
> if (blk_queue_is_zoned(q) && !q->elevator)
>
> In other words, auto-hibern8 is enabled only if the above condition false.
>
> However, the during a normal operation, the ufshcd_auto_hibern8_update() may not get called at all, and auto-hibern8 can be enabled in SDB mode as part of ufs init. Would that be a problem to have auto-hibern8 enabled without checking whether the above condition is false?
Hi Bao,
Probing the host controller happens as follows:
* The host controller probing function is called, e.g. ufs_qcom_probe().
* ufshcd_alloc_host() and ufshcd_init() are called directly or indirectly
by the host controller probe function.
* ufshcd_init() calls scsi_add_host() and async_schedule(ufshcd_async_scan, hba).
Once ufshcd_async_scan() is called asynchronously, it performs the
following actions:
* ufs_probe_hba() is called. This function calls ufshcd_link_startup()
indirectly. That function invokes the link_startup_notify vop if it has
been defined. Some host drivers set hba->ahit from inside that vop.
* scsi_scan_host() is called and calls scsi_alloc_sdev() indirectly.
* scsi_alloc_sdev() calls blk_mq_init_queue() and shost->hostt->slave_alloc().
* scsi_add_lun() calls sdev->host->hostt->slave_configure().
* scsi_add_lun() calls scsi_sysfs_add_sdev(). This indirectly causes sd_probe()
to be called asynchronously.
* sd_probe() calls sd_revalidate_disk(). This results in an indirect call of
disk_set_zoned(). Additionally, the ELEVATOR_F_ZBD_SEQ_WRITE flag is set by
the sd driver if q->limits.driver_preserves_write_order has not been set.
* Still from inside sd_probe(), device_add_disk() is called and a scheduler
is selected based on the value of q->required_elevator_features.
There are two ways in which host drivers initialize auto-hibernation:
* Either by setting hba->ahit before ufshcd_init() is called.
* Or by calling ufshcd_auto_hibern8_update() from the link_startup_notify
vop.
I think the above shows that the zoned model is queried *after* the above methods
for enabling auto-hibernation have completed and hence that blk_queue_is_zoned()
evaluates to false if a host driver calls ufshcd_auto_hibern8_update() from
inside the link_startup_notify vop.
If auto-hibernation is enabled from inside the host driver then that should
happen before sd_probe() is called. sd_probe() will use the value of
q->limits.driver_preserves_write_order that has been set by
ufshcd_slave_configure() so there is no risk for inconsistencies between the
auto-hibernation configuration and the request queue configuration.
Bart.
prev parent reply other threads:[~2023-08-18 17:57 UTC|newest]
Thread overview: 46+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-08-16 19:53 [PATCH v9 00/17] Improve performance for zoned UFS devices Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 01/17] block: Introduce more member variables related to zone write locking Bart Van Assche
2023-08-17 11:00 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 02/17] block: Only use write locking if necessary Bart Van Assche
2023-08-17 11:01 ` Damien Le Moal
2023-08-17 14:21 ` Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 03/17] block/mq-deadline: Only use zone " Bart Van Assche
2023-08-17 11:02 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 04/17] scsi: core: Call .eh_prepare_resubmit() before resubmitting Bart Van Assche
2023-08-17 11:10 ` Damien Le Moal
2023-08-17 14:26 ` Bart Van Assche
2023-08-18 2:38 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 05/17] scsi: core: Add unit tests for scsi_call_prepare_resubmit() Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 06/17] scsi: sd: Sort commands by LBA before resubmitting Bart Van Assche
2023-08-17 11:13 ` Damien Le Moal
2023-08-17 14:34 ` Bart Van Assche
2023-08-18 2:41 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 07/17] scsi: core: Retry unaligned zoned writes Bart Van Assche
2023-08-17 11:21 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 08/17] scsi: sd_zbc: Only require an I/O scheduler if needed Bart Van Assche
2023-08-17 11:21 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 09/17] scsi: scsi_debug: Support disabling zone write locking Bart Van Assche
2023-08-17 11:25 ` Damien Le Moal
2023-08-17 14:35 ` Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 10/17] scsi: scsi_debug: Support injecting unaligned write errors Bart Van Assche
2023-08-17 11:29 ` Damien Le Moal
2023-08-16 19:53 ` [PATCH v9 11/17] scsi: ufs: hisi: Rework the code that disables auto-hibernation Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 12/17] scsi: ufs: mediatek: Rework the code for disabling auto-hibernation Bart Van Assche
2023-08-17 18:40 ` Bao D. Nguyen
2023-08-17 19:13 ` Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 13/17] scsi: ufs: sprd: " Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 14/17] scsi: ufs: Rename ufshcd_auto_hibern8_enable() and make it static Bart Van Assche
2023-08-17 18:49 ` Bao D. Nguyen
2023-08-17 19:16 ` Bart Van Assche
2023-08-17 21:48 ` Bao D. Nguyen
2023-08-16 19:53 ` [PATCH v9 15/17] scsi: ufs: Simplify ufshcd_auto_hibern8_update() Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 16/17] scsi: ufs: Forbid auto-hibernation without I/O scheduler Bart Van Assche
2023-08-17 18:50 ` Bao D. Nguyen
2023-08-17 19:18 ` Bart Van Assche
2023-08-16 19:53 ` [PATCH v9 17/17] scsi: ufs: Inform the block layer about write ordering Bart Van Assche
2023-08-17 19:00 ` Bao D. Nguyen
2023-08-17 19:34 ` Bart Van Assche
2023-08-17 21:47 ` Bao D. Nguyen
2023-08-17 22:05 ` Bart Van Assche
2023-08-18 0:19 ` Bao D. Nguyen
2023-08-18 17:56 ` Bart Van Assche [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=06d34c2f-a89e-3134-d78f-3c071aa490cf@acm.org \
--to=bvanassche@acm.org \
--cc=Arthur.Simchaev@wdc.com \
--cc=avri.altman@wdc.com \
--cc=axboe@kernel.dk \
--cc=beanhuo@micron.com \
--cc=dlemoal@kernel.org \
--cc=hch@lst.de \
--cc=jejb@linux.ibm.com \
--cc=linux-block@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=martin.petersen@oracle.com \
--cc=ming.lei@redhat.com \
--cc=quic_asutoshd@quicinc.com \
--cc=quic_cang@quicinc.com \
--cc=quic_nguyenb@quicinc.com \
--cc=stanley.chu@mediatek.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox