| 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
| ||
|
Message-ID: <66d5fc8d-15ac-290c-8bb9-67900fe125a1@codeaurora.org>
Date: Thu, 21 Oct 2021 15:11:26 +0530
From: Maulik Shah <mkshah@...eaurora.org>
To: Shawn Guo <shawn.guo@...aro.org>
Cc: swboyd@...omium.org, mka@...omium.org, evgreen@...omium.org,
bjorn.andersson@...aro.org, linux-arm-msm@...r.kernel.org,
linux-kernel@...r.kernel.org, agross@...nel.org,
dianders@...omium.org, linux@...ck-us.net, rnayak@...eaurora.org,
lsrao@...eaurora.org,
Mahesh Sivasubramanian <msivasub@...eaurora.org>,
Lina Iyer <ilina@...eaurora.org>
Subject: Re: [PATCH v12 2/5] soc: qcom: Add Sleep stats driver
Hi Shawn,
On 10/21/2021 7:27 AM, Shawn Guo wrote:
> On Tue, Oct 19, 2021 at 06:16:57PM +0530, Maulik Shah wrote:
>> Hi Shawn,
>>
>> On 10/19/2021 3:17 PM, Shawn Guo wrote:
>>> On Mon, Oct 18, 2021 at 07:45:30PM +0530, Maulik Shah wrote:
>>>>>> +static void qcom_create_soc_sleep_stat_files(struct dentry *root, void __iomem *reg,
>>>>>> + struct stats_data *d,
>>>>>> + const struct stats_config *config)
>>>>>> +{
>>>>>> + char stat_type[sizeof(u32) + 1] = {0};
>>>>>> + size_t stats_offset = config->stats_offset;
>>>>>> + u32 offset = 0, type;
>>>>>> + int i, j;
>>>>>> +
>>>>>> + /*
>>>>>> + * On RPM targets, stats offset location is dynamic and changes from target
>>>>>> + * to target and sometimes from build to build for same target.
>>>>>> + *
>>>>>> + * In such cases the dynamic address is present at 0x14 offset from base
>>>>>> + * address in devicetree. The last 16bits indicates the stats_offset.
>>>>>> + */
>>>>>> + if (config->dynamic_offset) {
>>>>>> + stats_offset = readl(reg + RPM_DYNAMIC_ADDR);
>>>>>> + stats_offset &= RPM_DYNAMIC_ADDR_MASK;
>>>>>> + }
>>>>>> +
>>>>>> + for (i = 0; i < config->num_records; i++) {
>>>>>> + d[i].base = reg + offset + stats_offset;
>>>>>> +
>>>>>> + /*
>>>>>> + * Read the low power mode name and create debugfs file for it.
>>>>>> + * The names read could be of below,
>>>>>> + * (may change depending on low power mode supported).
>>>>>> + * For rpmh-sleep-stats: "aosd", "cxsd" and "ddr".
>>>>>> + * For rpm-sleep-stats: "vmin" and "vlow".
>>>>>
>>>>> It reports 'vmin' and 'xosd' on MSM8939, 'vmin' and 'vlow' on SDM660.
>>>>> I know that 'vmin' is VDD Minimization mode, and 'xosd' is XO Shutdown
>>>>> mode. But I'm not sure about 'vlow' mode. Could you share some
>>>>> information regarding what this low power mode is, and how it differs
>>>>> from 'vmin' and 'xosd'?
>>>>
>>>> vlow and xosd are same.
>>>> vmin is xosd plus voltage minimization of chip, memory rails.
>>>
>>> Thanks much for the info, Maulik!
>>>
>>> I'm running your driver on qcm2290 and trying to reach vlow mode.
>>>
>>> # cat /sys/kernel/debug/qcom_sleep_stats/vlow
>>> Count: 0
>>> Last Entered At: 0
>>> Last Exited At: 0
>>> Accumulated Duration: 0
>>> Client Votes: 0x81
>>> # echo mem > /sys/power/state
>>> [ 551.446603] PM: suspend entry (s2idle)
>>> [ 551.450948] Filesystems sync: 0.000 seconds
>>> [ 551.462828] Freezing user space processes ... (elapsed 0.002 seconds) done.
>>> [ 551.472276] OOM killer disabled.
>>> [ 551.475556] Freezing remaining freezable tasks ... (elapsed 0.001 seconds) done.
>>> [ 551.484461] printk: Suspending console(s) (use no_console_suspend to debug)
>>> [ 551.561280] OOM killer enabled.
>>> [ 551.564461] Restarting tasks ... done.
>>> [ 551.569652] PM: suspend exit
>>> # cat /sys/kernel/debug/qcom_sleep_stats/vlow
>>> Count: 0
>>> Last Entered At: 0
>>> Last Exited At: 0
>>> Accumulated Duration: 0
>>> Client Votes: 0x818081
>>>
>>> The count doesn't increases along with suspend/resume cycle at the
>>> moment. But as you can see, 'Client Votes' field changes. If possible,
>>> could you shed some light on what this means?
>>
>> The count will increase only when all the subsystems (APSS/Modem,etc) are in
>> power down mode and finally RPM decides to turn off xo clock.
>>
>>>
>>> As the comparison, I'm also running the downstream 'rpm_master_stats'
>>> driver in the same kernel, and the 'xo_count' field of APSS does
>>> increase along with suspend/resume cycle. May I ask some advices what
>>> I'm possibly missing and thus getting different result between 'vlow'
>>> and 'rpm_master_stats' report?
>>
>> The vlow is a SoC level state whereas the rpm master stats indicate
>> individual subsystem state. Since you are running suspend-resume the APSS is
>> going to sleep so you see xo_count incremented for it but for MPSS i see it
>> does not increase (modem is not entering to low power mode). similarly for
>> ADSP/CDSP it does not increment. if all of these subsystems goes to power
>> down and then there is sufficient sleep time for the SoC then you may see
>> vlow/vmin incrementing.
>>
>> Hope this clarifies.
>
> Thanks Maulik! It's very helpful. I have a couple of further
> questions, if you do not mind.
>
> 1. We can understand most of vlow/vmin output. But could you help
> decode 'Client Votes'? It looks like the bits are shifting along
> with suspend/resume cycle.
>
Client votes are history of client votes on xo. Each byte represents xo
clock votes from subsystems.
> 2. In the rpm_master_stats output below, I know masters (processors)
> APSS, MPSS, ADSP and CDSP, but not really sure what TZ is. If it's
> TrustZone, shouldn't it covered by APSS?
Yes it indicates for Trustzone but should be covered by APSS.
Thanks,
Maulik
>
> Thanks for sharing your insights!
>
> Shawn
>
>>> # cat /sys/kernel/debug/rpm_master_stats
>>> APSS
>>> shutdown_req:0x37EA3CC74
>>> wakeup_ind:0x0
>>> bringup_req:0x37F041958
>>> bringup_ack:0x37F042D54
>>> xo_last_entered_at:0x286FF36AC
>>> xo_last_exited_at:0x28AF94178
>>> xo_accumulated_duration:0x3EDD55B
>>> last_sleep_transition_duration:0x122f
>>> last_wake_transition_duration:0x11f8
>>> xo_count:0x1
>>> wakeup_reason:0x0
>>> numshutdowns:0x641
>>> active_cores:0x1
>>> core0
>>> MPSS
>>> shutdown_req:0x0
>>> wakeup_ind:0x0
>>> bringup_req:0x0
>>> bringup_ack:0x0
>>> xo_last_entered_at:0x0
>>> xo_last_exited_at:0x0
>>> xo_accumulated_duration:0x0
>>> last_sleep_transition_duration:0x0
>>> last_wake_transition_duration:0x0
>>> xo_count:0x0
>>> wakeup_reason:0x0
>>> numshutdowns:0x0
>>> active_cores:0x1
>>> core0
>>> ADSP
>>> shutdown_req:0x0
>>> wakeup_ind:0x0
>>> bringup_req:0x0
>>> bringup_ack:0x0
>>> xo_last_entered_at:0x0
>>> xo_last_exited_at:0x0
>>> xo_accumulated_duration:0x0
>>> last_sleep_transition_duration:0x0
>>> last_wake_transition_duration:0x0
>>> xo_count:0x0
>>> wakeup_reason:0x0
>>> numshutdowns:0x0
>>> active_cores:0x1
>>> core0
>>> CDSP
>>> shutdown_req:0x0
>>> wakeup_ind:0x0
>>> bringup_req:0x0
>>> bringup_ack:0x0
>>> xo_last_entered_at:0x0
>>> xo_last_exited_at:0x0
>>> xo_accumulated_duration:0x0
>>> last_sleep_transition_duration:0x0
>>> last_wake_transition_duration:0x0
>>> xo_count:0x0
>>> wakeup_reason:0x0
>>> numshutdowns:0x0
>>> active_cores:0x0
>>> TZ
>>> shutdown_req:0x0
>>> wakeup_ind:0x0
>>> bringup_req:0x0
>>> bringup_ack:0x0
>>> xo_last_entered_at:0x0
>>> xo_last_exited_at:0x0
>>> xo_accumulated_duration:0x0
>>> last_sleep_transition_duration:0x0
>>> last_wake_transition_duration:0x0
>>> xo_count:0x0
>>> wakeup_reason:0x0
>>> numshutdowns:0x0
>>> active_cores:0x0
--
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a
member of Code Aurora Forum, hosted by The Linux Foundation
Powered by blists - more mailing lists