From: Tom Rix <trix@redhat.com>
To: Luca Ceresoli <luca@lucaceresoli.net>
Cc: linux-fpga@vger.kernel.org, Moritz Fischer <mdf@kernel.org>,
Michal Simek <michal.simek@xilinx.com>,
linux-arm-kernel@lists.infradead.org,
linux-kernel@vger.kernel.org, Anatolij Gustschin <agust@denx.de>
Subject: Re: [PATCH v2 3/5] fpga manager: xilinx-spi: rework write_complete loop implementation
Date: Fri, 28 Aug 2020 05:34:02 -0700 [thread overview]
Message-ID: <67ad9e7e-b07a-52e7-f82d-b7f8fac28d8f@redhat.com> (raw)
In-Reply-To: <913cb7ea-3502-b3cd-3ec9-af60d63129c9@lucaceresoli.net>
On 8/27/20 11:38 PM, Luca Ceresoli wrote:
> Hi Tom,
>
> On 27/08/20 21:34, Tom Rix wrote:
>> On 8/27/20 12:26 PM, Luca Ceresoli wrote:
>>> Hi Tom,
>>>
>>> thanks for the prompt feedback!
>>>
>>> On 27/08/20 20:59, Tom Rix wrote:
>>>> On 8/27/20 7:32 AM, Luca Ceresoli wrote:
>>>>> In preparation to add error checking for gpiod_get_value(), rework
>>>>> the loop to avoid the duplication of these lines:
>>>>>
>>>>> if (gpiod_get_value(conf->done))
>>>>> return xilinx_spi_apply_cclk_cycles(conf);
>>>>>
>>>>> There is little advantage in this rework with current code. However
>>>>> error checking will expand these two lines to five, making code
>>>>> duplication more annoying.
>>>>>
>>>>> Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
>>>>>
>>>>> ---
>>>>>
>>>>> This patch is new in v2
>>>>> ---
>>>>> drivers/fpga/xilinx-spi.c | 15 ++++++---------
>>>>> 1 file changed, 6 insertions(+), 9 deletions(-)
>>>>>
>>>>> diff --git a/drivers/fpga/xilinx-spi.c b/drivers/fpga/xilinx-spi.c
>>>>> index 01f494172379..cfc933d70f52 100644
>>>>> --- a/drivers/fpga/xilinx-spi.c
>>>>> +++ b/drivers/fpga/xilinx-spi.c
>>>>> @@ -151,22 +151,19 @@ static int xilinx_spi_write_complete(struct fpga_manager *mgr,
>>>>> struct fpga_image_info *info)
>>>>> {
>>>>> struct xilinx_spi_conf *conf = mgr->priv;
>>>>> - unsigned long timeout;
>>>>> + unsigned long timeout = jiffies + usecs_to_jiffies(info->config_complete_timeout_us);
>>>>> int ret;
>>>>>
>>>>> - if (gpiod_get_value(conf->done))
>>>>> - return xilinx_spi_apply_cclk_cycles(conf);
>>>>> -
>>>>> - timeout = jiffies + usecs_to_jiffies(info->config_complete_timeout_us);
>>>>> + while (true) {
>>>>> + if (gpiod_get_value(conf->done))
>>>>> + return xilinx_spi_apply_cclk_cycles(conf);
>>>>>
>>>>> - while (time_before(jiffies, timeout)) {
>>>>> + if (time_after(jiffies, timeout))
>>>>> + break;
>>>>>
>>>>> ret = xilinx_spi_apply_cclk_cycles(conf);
>>>>> if (ret)
>>>>> return ret;
>>>>> -
>>>>> - if (gpiod_get_value(conf->done))
>>>>> - return xilinx_spi_apply_cclk_cycles(conf);
>>>>> }
>>>> Do you need another
>>>>
>>>> if (gpiod_get_value(conf->done))
>>>> return xilinx_spi_apply_cclk_cycles(conf);
>>>>
>>>> here to cover the chance of sleeping in the loop ?
>>> If I got your question correctly: if we get here it's because of a
>>> timeout, thus programming has failed (DONE didn't come up after some
>>> time), and checking it one more here seems pointless.
>> It may not be pointless, if this routine sleeps because it was scheduled out, when it wakes up a lot of time happened. You will see this as a timeout but the state may be good. Another, final check at the end will cover this case.
> Oh, now I got your point! Yes, there is this risk, and it exists in
> current code as well but with a smaller risk window. Unrolling the
> current and new loop code they behave the same except for the position
> of the timeout computation (after vs before the first 'if (done) return'
> group).
>
> I think this reimplementation is sleep-safe, check for GPIO errors and
> also avoid code duplication:
>
> static int xilinx_spi_write_complete(struct fpga_manager *mgr,
> struct fpga_image_info *info)
> {
> struct xilinx_spi_conf *conf = mgr->priv;
> unsigned long timeout = jiffies +
> usecs_to_jiffies(info->config_complete_timeout_us);
> bool expired;
> int done;
> int ret;
>
> while (!expired) {
> expired = time_after(jiffies, timeout);
>
> done = get_done_gpio(mgr);
> if (done < 0)
> return done;
>
> ret = xilinx_spi_apply_cclk_cycles(conf);
> if (ret)
> return ret;
>
> if (done)
> return 0;
> }
>
> dev_err(&mgr->dev, "Timeout after config data transfer\n");
>
> return -ETIMEDOUT;
> }
>
> A key point is to assess all the status (expired and done variables)
> before taking any action based on it. Then we can unconditionally apply
> 8 cclk cycles before even checking the actual DONE value, so that we
> always do that after DONE has been seen asserted.
>
> Does it look good?
Yes. Thanks for the extra work.
Tom
>
next prev parent reply other threads:[~2020-08-28 12:34 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-27 14:32 [PATCH v2 1/5] fpga manager: xilinx-spi: remove stray comment Luca Ceresoli
2020-08-27 14:32 ` [PATCH v2 2/5] fpga manager: xilinx-spi: remove final dot from dev_err() strings Luca Ceresoli
2020-08-27 14:32 ` [PATCH v2 3/5] fpga manager: xilinx-spi: rework write_complete loop implementation Luca Ceresoli
2020-08-27 18:59 ` Tom Rix
2020-08-27 19:26 ` Luca Ceresoli
[not found] ` <27bbb896-fedf-6a3a-7220-5c57239a3b87@redhat.com>
2020-08-28 6:38 ` Luca Ceresoli
2020-08-28 12:34 ` Tom Rix [this message]
2020-08-27 14:32 ` [PATCH v2 4/5] fpga manager: xilinx-spi: add error checking after gpiod_get_value() Luca Ceresoli
2020-08-27 19:04 ` Tom Rix
2020-08-27 14:32 ` [PATCH v2 5/5] fpga manager: xilinx-spi: provide better diagnostics on programming failure Luca Ceresoli
2020-08-27 19:09 ` Tom Rix
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=67ad9e7e-b07a-52e7-f82d-b7f8fac28d8f@redhat.com \
--to=trix@redhat.com \
--cc=agust@denx.de \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-fpga@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=luca@lucaceresoli.net \
--cc=mdf@kernel.org \
--cc=michal.simek@xilinx.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;
as well as URLs for NNTP newsgroup(s).