linux-kernel - Re: [PATCH 1/2] drivers: qcom: rpmh-rsc: simplify TCS locking

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 22 Jul 2019 10:20:03 -0600
From:   Lina Iyer <ilina@...eaurora.org>
To:     Stephen Boyd <swboyd@...omium.org>
Cc:     andy.gross@...aro.org, bjorn.andersson@...aro.org,
        linux-arm-msm@...r.kernel.org, linux-soc@...r.kernel.org,
        rnayak@...eaurora.org, linux-kernel@...r.kernel.org,
        linux-pm@...r.kernel.org, dianders@...omium.org,
        mkshah@...eaurora.org, "Raju P.L.S.S.S.N" <rplsssn@...eaurora.org>
Subject: Re: [PATCH 1/2] drivers: qcom: rpmh-rsc: simplify TCS locking

On Fri, Jul 19 2019 at 12:20 -0600, Stephen Boyd wrote:
>Quoting Lina Iyer (2019-07-01 08:29:06)
>> From: "Raju P.L.S.S.S.N" <rplsssn@...eaurora.org>
>>
>> tcs->lock was introduced to serialize access with in TCS group. But
>> even without tcs->lock, drv->lock is serving the same purpose. So
>> use a single drv->lock.
>
>Isn't the downside now that we're going to be serializing access to the
>different TCSes when two are being written in parallel or waited on? I
>thought that was the whole point of splitting the lock into a TCS lock
>and a general "driver" lock that protects the global driver state vs.
>the specific TCS state.
>
Yes but we were holding the drv->lock as well as tcs->lock for the most
critical of the path anyways (writing to TCS). The added complexity
doesn't seem to help reduce the latency that it expected to reduce.
>>
>> Other optimizations include -
>>  - Remove locking around clear_bit() in IRQ handler. clear_bit() is
>>    atomic.
>>  - Remove redundant read of TCS registers.
>>  - Use spin_lock instead of _irq variants as the locks are not held
>>    in interrupt context.
>
>Can you please split this patch up into 3 or 4 different patches? I'm
>not sure why any of these patches are marked with Fixes either. It's an
>optimization patch, not a fix patch, unless the optimization is really
>large somehow?
>
Okay. I will try that.
>>
>> Fixes: 658628 ("drivers: qcom: rpmh-rsc: add RPMH controller for QCOM
>> SoCs")
>> Signed-off-by: Raju P.L.S.S.S.N <rplsssn@...eaurora.org>
>> Signed-off-by: Lina Iyer <ilina@...eaurora.org>
>> ---
>>  drivers/soc/qcom/rpmh-internal.h |  2 --
>>  drivers/soc/qcom/rpmh-rsc.c      | 37 +++++++++++---------------------
>>  drivers/soc/qcom/rpmh.c          | 20 +++++++----------
>>  3 files changed, 21 insertions(+), 38 deletions(-)
>>
>> diff --git a/drivers/soc/qcom/rpmh-internal.h b/drivers/soc/qcom/rpmh-internal.h
>> index a7bbbb67991c..969d5030860e 100644
>> --- a/drivers/soc/qcom/rpmh-internal.h
>> +++ b/drivers/soc/qcom/rpmh-internal.h
>> diff --git a/drivers/soc/qcom/rpmh-rsc.c b/drivers/soc/qcom/rpmh-rsc.c
>> index e278fc11fe5c..92461311aef3 100644
>> --- a/drivers/soc/qcom/rpmh-rsc.c
>> +++ b/drivers/soc/qcom/rpmh-rsc.c
>> @@ -93,8 +93,7 @@ static void write_tcs_reg_sync(struct rsc_drv *drv, int reg, int tcs_id,
>>
>>  static bool tcs_is_free(struct rsc_drv *drv, int tcs_id)
>>  {
>> -       return !test_bit(tcs_id, drv->tcs_in_use) &&
>> -              read_tcs_reg(drv, RSC_DRV_STATUS, tcs_id, 0);
>> +       return !test_bit(tcs_id, drv->tcs_in_use);
>
>This can be a diffedjusted rent patch. Why is reading the tcs register
>redundant? Please put that information in the commit text.
>
The tcs_in_use, is adjusted along with the DRV_STS and reading the
tcs_in_use should be enough.

Thanks for your review Stephen.

--Lina