Received-SPF: pass (google.com: domain of linux-wireless-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII;
 format=flowed
Content-Transfer-Encoding: 7bit
Date:   Tue, 25 Aug 2020 11:41:03 +0800
From:   Wen Gong <wgong@codeaurora.org>
To:     Krishna Chaitanya <chaitanya.mgit@gmail.com>
Cc:     ath10k <ath10k@lists.infradead.org>,
        linux-wireless <linux-wireless@vger.kernel.org>
Subject: Re: [PATCH v2] ath10k: add flag to protect napi operation to avoid
 dead loop hang for SDIO
In-Reply-To: <CABPxzY+Fc22ke1ygRnDs0Uds0a7n=792QN8ALXciDrnx9SQuqQ@mail.gmail.com>
References: <1598243612-4627-1-git-send-email-wgong@codeaurora.org>
 <CABPxzYLjys+cXXRM5J680ZOs+6VrYt=_3rWv-gqkCod=-A1VrA@mail.gmail.com>
 <fd98989a87f2a50655dc95bdcd535c0d@codeaurora.org>
 <CABPxzYKvPwtQwxMfRcv9jT+d92ErhYGR91SKBH86T3Rd2QH9Qg@mail.gmail.com>
 <2d6362ce85956d0f7df2e596b89a7028@codeaurora.org>
 <CABPxzY+Fc22ke1ygRnDs0Uds0a7n=792QN8ALXciDrnx9SQuqQ@mail.gmail.com>
Message-ID: <663fbb682fd8df555b2964d7d86b8eca@codeaurora.org>
User-Agent: Roundcube Webmail/1.3.9
Sender: linux-wireless-owner@vger.kernel.org
Precedence: bulk

On 2020-08-24 19:15, Krishna Chaitanya wrote:
> On Mon, Aug 24, 2020 at 4:15 PM Wen Gong <wgong@codeaurora.org> wrote:
>> 
>> On 2020-08-24 18:03, Krishna Chaitanya wrote:
>> > On Mon, Aug 24, 2020 at 3:10 PM Wen Gong <wgong@codeaurora.org> wrote:
>> >>
>> >> On 2020-08-24 16:35, Krishna Chaitanya wrote:
>> >> > On Mon, Aug 24, 2020 at 10:03 AM Wen Gong <wgong@codeaurora.org> wrote:
>> >> >>
>> >> >> It happened "Kernel panic - not syncing: hung_task: blocked tasks"
>> >> >> when
>> >> >> test simulate crash and ifconfig down/rmmod meanwhile.
>> >> >>
>> >> ...
>> >> >>
>> >> >>  #ifdef CONFIG_PM
>> >> > Even though your DUT is SDIO based we should be doing this in general
>> >> > for all, no?
>> >> > core_restart + hif_stop is common to all.
>> >> this patch does not have core_restart.
>> > I was referring to the combination which is causing the issue.
>> >
>> >> I dit not hit the issue for others bus(PCIe,SNOC...), so I can not
>> >> change them with a
>> >> assumption they also have this issue.
>> > But that doesn't make sense, the combination is being hit for others
>> > also.
>> > (they should also endup calling napi_disable twice?) or they are using
>> > some other check to avoid this (doesn't appear so from a quick look at
>> > the
>> > code).
>> Because I only use SDIO, I did not use others BUS, so I did not hit 
>> the
>> issue
>> on other BUS.
> I understand, my point was based on the description the issue looks 
> independent
> of the BUS type, so, the fix should also be generic. I understand that
> your testing
> is only focused on SDIO, but we should have a generic fix and probably 
> use
> communities help to get it tested rather than fixing SDIO only.
I checked the ath10k, only sdio.c, snoc.c, pci.c have used napi.
I think it can change to move the 
napi_synchronize/napi_disable/napi_enable from
sido.c/snoc.c/pci.c to ath10k_core.ko as below:
void ath10k_core_napi_enable(struct ath10k *ar)
{
        if (!ar->napi_enabled) {
                napi_enable(&ar->napi);
                ar->napi_enabled = true;
        }
}
EXPORT_SYMBOL(ath10k_core_napi_enable);

void ath10k_core_napi_disable_sync(struct ath10k *ar)
{
        if (ar->napi_enabled) {
                napi_synchronize(&ar->napi);
                napi_disable(&ar->napi);
                ar->napi_enabled = false;
        }
}
EXPORT_SYMBOL(ath10k_core_napi_disable_sync);

is it appropriate?
...