mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Paul Menzel <pmenzel@molgen.mpg.de>
To: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Cc: linux-ide@vger.kernel.org, linux-kernel@vger.kernel.org,
	Jeff Garzik <jeff@bloq.com>, Tejun Heo <tj@kernel.org>
Subject: ata: For what PHY was debounce delay introduced? (was: [PATCH v3 3/3] ahci: AMD A85 FCH (Hudson D4): Skip 200 ms debounce delay in `sata_link_resume()`)
Date: Tue, 4 Jan 2022 12:34:24 +0100	[thread overview]
Message-ID: <66edbfd4-063f-b995-0d15-982d365dd7d7@molgen.mpg.de> (raw)
In-Reply-To: <27da2f5c-ca6c-1d64-3d05-5453f11e298f@opensource.wdc.com>

[cc: +jeff, +tejun]

Dear Damien,


Am 04.01.22 um 10:08 schrieb Damien Le Moal:
> On 1/4/22 17:49, Paul Menzel wrote:
>> [cc: -dmitry, -guenter]

>> Am 04.01.22 um 09:36 schrieb Damien Le Moal:
>>> On 12/31/21 16:08, Paul Menzel wrote:
>>
>>>> Am 31.12.21 um 01:52 schrieb Damien Le Moal:
>>>>> On 12/30/21 20:08, Paul Menzel wrote:

[…]

Jeff, Tejun, in `drivers/ata/libata-sata.c` contains a 200 ms sleep 
delaying the boot noticeably for optimized setups, where Linux takes 
less than 500 ms to start:

```
int sata_link_resume(struct ata_link *link, const unsigned long *params,
                      unsigned long deadline)
{
[…]
         /*
          * Writes to SControl sometimes get ignored under certain
          * controllers (ata_piix SIDPR).  Make sure DET actually is
          * cleared.
          */
         do {
                 scontrol = (scontrol & 0x0f0) | 0x300;
                 if ((rc = sata_scr_write(link, SCR_CONTROL, scontrol)))
                         return rc;
                 /*
                  * Some PHYs react badly if SStatus is pounded
                  * immediately after resuming.  Delay 200ms before
                  * debouncing.
                  */
                 if (!(link->flags & ATA_LFLAG_NO_DB_DELAY))
                         ata_msleep(link->ap, 200);

                 /* is SControl restored correctly? */
                 if ((rc = sata_scr_read(link, SCR_CONTROL, &scontrol)))
                         return rc;
         } while ((scontrol & 0xf0f) != 0x300 && --tries);
[…]
}
```

>>> It would indeed be great to have the default as "no delay on resume" and
>>> add the delay only for chipsets that need it. However, it is unclear
>>> which chipset need the delay, right?
>>
>> Yes, it’s unclear for what chipset (PHY?) it was added, as the git
>> history i not available in the repository, and I have not found it yet.

I found the historical git archive [2], and Jeff’s commit 4effb658a0 
from October 2003 [3] with the commit message

> [libata] Merge Serial ATA core, and drivers for:
> 
> Intel ICH5 (production)
> ServerWorks / Apple K2 (beta)
> VIA (beta)
> Silicon Image 3112 (broken!)
> Various Promise (alpha/beta)

adds the code below:

```
void sata_phy_reset(struct ata_port *ap)
{
[…]
	/* wait for phy to become ready, if necessary */
	do {
		msleep(200);
		sstatus = scr_read(ap, SCR_STATUS);
		if ((sstatus & 0xf) != 1)
			break;
	} while (time_before(jiffies, timeout));
[…]
}
```

Later on Tejun refactored the code in commit d7bb4cc75759 ([PATCH] 
libata-hp-prep: implement sata_phy_debounce()) [4], and clarified the 
comment.

(Sorry, if I mis-analyzed anything.)

Jeff, Tejun, do you know, what chipsets/PHYs had trouble with being 
queried right away? Only ata_piix?

>>> So I think we are stuck with switching chipsets to "no delay" one by
>>> one by testing. Once the majority of drivers are converted, we can
>>> reverse the default to be "no delay" and mark untested drivers as
>>> needing the delay.
>>
>> For easy testing, a new CLI parameter to skip the delay might be handy.
> 
> You mean a sysfs attribute may be ?
> I am not sure it would help: on resume, the sysfs attributes would be
> recreated and get the default value, not a new one.

No, I mean a module parameter for `libata` like `ata_probe_timeout`. 
Then users could test it during boot, and, if there are no issue, tell 
us the ID.


Kind regards,

Paul


>>>> [1]: https://lore.kernel.org/linux-ide/20211227162658.11314-2-pmenzel@molgen.mpg.de/T/#m697d2121463a4c946730e6b83940e12d6d7e6700
[2]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/
[3]: 
https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=4effb658a0f800e159c29a2d881cac76c326087a
[4]: 
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=d7bb4cc7575929a60b0a718daa1bce87bea9a9cc

  reply	other threads:[~2022-01-04 11:34 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-12-29 16:11 [PATCH v3 1/3] PCI: Add device code for AMD FCH SATA Controller in AHCI mode Paul Menzel
2021-12-29 16:11 ` [PATCH v3 2/3] ahci: Use macro PCI_DEVICE_ID_AMD_HUDSON2_SATA_IDE Paul Menzel
2021-12-30  2:13   ` Damien Le Moal
2021-12-30 11:01     ` Paul Menzel
2021-12-31  0:50       ` Damien Le Moal
2021-12-30 13:35   ` Hannes Reinecke
2021-12-29 16:11 ` [PATCH v3 3/3] ahci: AMD A85 FCH (Hudson D4): Skip 200 ms debounce delay in `sata_link_resume()` Paul Menzel
2021-12-30  2:19   ` Damien Le Moal
2021-12-30 11:08     ` Paul Menzel
2021-12-31  0:52       ` Damien Le Moal
2021-12-31  7:08         ` Paul Menzel
2022-01-04  8:36           ` Damien Le Moal
2022-01-04  8:49             ` Paul Menzel
2022-01-04  9:08               ` Damien Le Moal
2022-01-04 11:34                 ` Paul Menzel [this message]
     [not found]                   ` <CAHy7=M10exWuVJtDVo+w36YKadY533GttzwrKxPnotHf8-JQnw@mail.gmail.com>
2022-01-07 21:30                     ` ata: For what PHY was debounce delay introduced? (was: [PATCH v3 3/3] ahci: AMD A85 FCH (Hudson D4): Skip 200 ms debounce delay in `sata_link_resume()`) Tejun Heo
2022-01-04  6:04   ` [PATCH v3 3/3] ahci: AMD A85 FCH (Hudson D4): Skip 200 ms debounce delay in `sata_link_resume()` Dmitry Torokhov
2021-12-30 19:21 ` [PATCH v3 1/3] PCI: Add device code for AMD FCH SATA Controller in AHCI mode Sergei Shtylyov

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=66edbfd4-063f-b995-0d15-982d365dd7d7@molgen.mpg.de \
    --to=pmenzel@molgen.mpg.de \
    --cc=damien.lemoal@opensource.wdc.com \
    --cc=jeff@bloq.com \
    --cc=linux-ide@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=tj@kernel.org \
    /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

all inboxes | Powered by JetHome®