Date:   Mon, 02 Nov 2020 22:05:07 +0900
From:   Tsuchiya Yuto <kitakar@...il.com>
To:     Amitkumar Karwar <amitkarwar@...il.com>,
        Ganapathi Bhat <ganapathi.bhat@....com>,
        Xinming Hu <huxinming820@...il.com>,
        Kalle Valo <kvalo@...eaurora.org>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>
Cc:     linux-wireless@...r.kernel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Maximilian Luz <luzmaximilian@...il.com>,
        Andy Shevchenko <andriy.shevchenko@...el.com>, verdre@...d.nl
Subject: Re: [RFC PATCH] mwifiex: pcie: use shutdown_sw()/reinit_sw() on
 suspend/resume

On Sat, 2020-10-31 at 00:27 +0900, Tsuchiya Yuto wrote:
> On Wed, 2020-10-28 at 23:27 +0900, Tsuchiya Yuto wrote:
>> On Microsoft Surface devices (PCIe-88W8897), there are issues with S0ix
>> achievement and AP scanning after suspend with the current Host Sleep
>> method.
>>
>> When using the Host Sleep method, it prevents the platform to reach S0ix
>> during suspend. Also, sometimes AP scanning won't work, resulting in
>> non-working wifi after suspend.
>>
>> To fix such issues, perform shutdown_sw()/reinit_sw() instead of Host
>> Sleep on suspend/resume.
>>
>> Signed-off-by: Tsuchiya Yuto <kitakar@...il.com>
>> ---
>> As a side effect, this patch disables wakeups (means that Wake-On-WLAN
>> can't be used anymore, if it was working before), and might also reset
>> some internal states.
>>
>> Of course it's the best to rather fix Host Sleep itself. But if it's
>> difficult, I'm afraid we have to go this way.
>>
>> I reused the contents of suspend()/resume() functions as much as possible,
>> and removed only the parts that are incompatible or redundant with
>> shutdown_sw()/reinit_sw().
>>
>> - Removed wait_for_completion() as redundant
>>   mwifiex_shutdown_sw() does this.
>> - Removed flush_workqueue() as incompatible
>>   Causes kernel crashing.
>> - Removed mwifiex_enable_wake()/mwifiex_disable_wake()
>>   as incompatible and redundant because the driver will be shut down
>>   instead of entering Host Sleep.
>>
>> I'm worried about why flush_workqueue() causes kernel crash with this
>> suspend method. Is it OK to just drop it? At least We Microsoft Surface
>> devices users used this method for about one month and haven't observed
>> any issues.
>>
>> Note that suspend() no longer checks if it's already suspended.
>> With the previous Host Sleep method, the check was done by looking at
>> adapter->hs_activated in mwifiex_enable_hs() [sta_ioctl.c], but not
>> MWIFIEX_IS_SUSPENDED. So, what the previous method checked was instead
>> Host Sleep state, not suspend itself.
>>
>> Therefore, there is no need to check the suspend state now.
>> Also removed comment for suspend state check at top of suspend()
>> accordingly.
>
> This patch depends on the following mwifiex_shutdown_sw() fix I sent
> separately.
>
> [PATCH 1/2] mwifiex: fix mwifiex_shutdown_sw() causing sw reset failure
> https://lore.kernel.org/linux-wireless/20201028142110.18144-2-kitakar@gmail.com/

The AP scanning issue with Host Sleep is now difficult to reproduce on
v5.10-rc2. It might be already gone but not yet so sure.

Here are the details about AP scanning issue with Host Sleep for the
record (as of Apr 2020):

When using Host Sleep on suspend, after resuming from suspend, it
(sometimes) can't connect to APs because it fails to scan APs. When
I set debug_mask to 0xffffffff, I noticed that scanning is being blocked
with this message:

    kern  :info  : [99952.621609] mwifiex_pcie 0000:03:00.0: info: received scan request on mlan0
    kern  :info  : [99952.621613] mwifiex_pcie 0000:03:00.0: cmd: Scan already in process..

What is worse, when this issue happened, the subsequent suspend
(sometimes) fails with the following message:

    kern  :info  : [101844.423427] mwifiex_pcie 0000:03:00.0: hs_activate_wait_q terminated
    kern  :info  : [101844.423433] mwifiex_pcie 0000:03:00.0: cmd: failed to suspend
    kern  :err   : [101844.423446] PM: pci_pm_suspend(): mwifiex_pcie_suspend+0x0/0xd0 [mwifiex_pcie] returns -14
    kern  :err   : [101844.423453] PM: dpm_run_callback(): pci_pm_suspend+0x0/0x160 returns -14
    kern  :err   : [101844.423466] PM: Device 0000:03:00.0 failed to suspend async: error -14
    kern  :debug : [101844.423525] PM: suspend of devices aborted after 10064.914 msecs
    kern  :debug : [101844.423529] PM: start suspend of devices aborted after 10065.318 msecs
    kern  :err   : [101844.423531] PM: Some devices failed to suspend, or early wake event detected

The message is from the following code in mwifiex_cfg80211_scan()
[cfg80211.c].

    /* Block scan request if scan operation or scan cleanup when interface
     * is disabled is in process
     */
    if (priv->scan_request || priv->scan_aborting) {
    	mwifiex_dbg(priv->adapter, WARN,
    		    "cmd: Scan already in process..\n");
    	return -EBUSY;
    }

Further print debugging showed that scan_request was not true but
scan_aborting was true. And the scan_aborting was set by mwifiex_close()
[main.c].

Regarding the S0ix achievement, I don't have any idea how I can fix it
with the Host Sleep method. So, I sent this patch. Any suggestions for
fixing it with Host Sleep are welcome.

If I understand correctly, the mwifiex card is in fully working state
in terms of PCIe. This prevents the platform from going into S0ix state?

>>  drivers/net/wireless/marvell/mwifiex/pcie.c | 29 +++++++--------------
>>  1 file changed, 10 insertions(+), 19 deletions(-)
>>
>> diff --git a/drivers/net/wireless/marvell/mwifiex/pcie.c b/drivers/net/wireless/marvell/mwifiex/pcie.c
>> index 6a10ff0377a24..3b5c614def2f5 100644
>> --- a/drivers/net/wireless/marvell/mwifiex/pcie.c
>> +++ b/drivers/net/wireless/marvell/mwifiex/pcie.c
>> @@ -293,8 +293,7 @@ static bool mwifiex_pcie_ok_to_access_hw(struct mwifiex_adapter *adapter)
>>   * registered functions must have drivers with suspend and resume
>>   * methods. Failing that the kernel simply removes the whole card.
>>   *
>> - * If already not suspended, this function allocates and sends a host
>> - * sleep activate request to the firmware and turns off the traffic.
>> + * This function shuts down the adapter.
>>   */
>>  static int mwifiex_pcie_suspend(struct device *dev)
>>  {
>> @@ -302,31 +301,21 @@ static int mwifiex_pcie_suspend(struct device *dev)
>>  	struct pcie_service_card *card = dev_get_drvdata(dev);
>>  
>>  
>> -	/* Might still be loading firmware */
>> -	wait_for_completion(&card->fw_done);
>> -
>>  	adapter = card->adapter;
>>  	if (!adapter) {
>>  		dev_err(dev, "adapter is not valid\n");
>>  		return 0;
>>  	}
>>  
>> -	mwifiex_enable_wake(adapter);
>> -
>> -	/* Enable the Host Sleep */
>> -	if (!mwifiex_enable_hs(adapter)) {
>> +	/* Shut down SW */
>> +	if (mwifiex_shutdown_sw(adapter)) {
>>  		mwifiex_dbg(adapter, ERROR,
>>  			    "cmd: failed to suspend\n");
>> -		clear_bit(MWIFIEX_IS_HS_ENABLING, &adapter->work_flags);
>> -		mwifiex_disable_wake(adapter);
>>  		return -EFAULT;
>>  	}
>>  
>> -	flush_workqueue(adapter->workqueue);
>> -
>>  	/* Indicate device suspended */
>>  	set_bit(MWIFIEX_IS_SUSPENDED, &adapter->work_flags);
>> -	clear_bit(MWIFIEX_IS_HS_ENABLING, &adapter->work_flags);
>>  
>>  	return 0;
>>  }
>> @@ -336,13 +325,13 @@ static int mwifiex_pcie_suspend(struct device *dev)
>>   * registered functions must have drivers with suspend and resume
>>   * methods. Failing that the kernel simply removes the whole card.
>>   *
>> - * If already not resumed, this function turns on the traffic and
>> - * sends a host sleep cancel request to the firmware.
>> + * If already not resumed, this function reinits the adapter.
>>   */
>>  static int mwifiex_pcie_resume(struct device *dev)
>>  {
>>  	struct mwifiex_adapter *adapter;
>>  	struct pcie_service_card *card = dev_get_drvdata(dev);
>> +	int ret;
>>  
>>  
>>  	if (!card->adapter) {
>> @@ -360,9 +349,11 @@ static int mwifiex_pcie_resume(struct device *dev)
>>  
>>  	clear_bit(MWIFIEX_IS_SUSPENDED, &adapter->work_flags);
>>  
>> -	mwifiex_cancel_hs(mwifiex_get_priv(adapter, MWIFIEX_BSS_ROLE_STA),
>> -			  MWIFIEX_ASYNC_CMD);
>> -	mwifiex_disable_wake(adapter);
>> +	ret = mwifiex_reinit_sw(adapter);
>> +	if (ret)
>> +		dev_err(dev, "reinit failed: %d\n", ret);
>> +	else
>> +		mwifiex_dbg(adapter, INFO, "%s, successful\n", __func__);
>>  
>>  	return 0;
>>  }
>
>

Powered by blists - more mailing lists