Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 2/3] mwifiex: support sysfs initiated device coredump
To:     Brian Norris <briannorris@chromium.org>
References: <1519210220-22437-1-git-send-email-arend.vanspriel@broadcom.com>
 <1519210220-22437-3-git-send-email-arend.vanspriel@broadcom.com>
 <20180221225903.GA42395@rodete-desktop-imager.corp.google.com>
 <5A8EB4F4.2030103@broadcom.com>
 <20180222193547.GA78462@rodete-desktop-imager.corp.google.com>
Cc:     Kalle Valo <kvalo@codeaurora.org>,
        Marcel Holtmann <marcel@holtmann.org>,
        linux-wireless@vger.kernel.org, linux-bluetooth@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Johannes Berg <johannes@sipsolutions.net>
From:   Arend van Spriel <arend.vanspriel@broadcom.com>
Message-ID: <5A8FEF68.5080900@broadcom.com>
Date:   Fri, 23 Feb 2018 11:39:36 +0100
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:38.0) Gecko/20100101
 Thunderbird/38.5.0
MIME-Version: 1.0
In-Reply-To: <20180222193547.GA78462@rodete-desktop-imager.corp.google.com>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

+ Johannes (for devcoredump documentation question).

On 2/22/2018 8:35 PM, Brian Norris wrote:
> Hi Arend,
>
> On Thu, Feb 22, 2018 at 01:17:56PM +0100, Arend van Spriel wrote:
>> On 2/21/2018 11:59 PM, Brian Norris wrote:
>>> On Wed, Feb 21, 2018 at 11:50:19AM +0100, Arend van Spriel wrote:
>>>> Since commit 3c47d19ff4dc ("drivers: base: add coredump driver ops")
>>>> it is possible to initiate a device coredump from user-space. This
>>>> patch adds support for it adding the .coredump() driver callback.
>>>> As there is no longer a need to initiate it through debugfs remove
>>>> that code.
>>>>
>>>> Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
>>>> ---
>>>>    drivers/net/wireless/marvell/mwifiex/debugfs.c | 31 +-------------------------
>>>>    drivers/net/wireless/marvell/mwifiex/pcie.c    | 19 ++++++++++++++--
>>>>    drivers/net/wireless/marvell/mwifiex/sdio.c    | 13 +++++++++++
>>>>    drivers/net/wireless/marvell/mwifiex/usb.c     | 14 ++++++++++++
>>>>    4 files changed, 45 insertions(+), 32 deletions(-)
>>>
>>> The documentation doesn't really say [1], but is the coredump supposed
>>> to happen synchronously? Because the mwifiex implementation is
>>> asynchronous, whereas it looks like the brcmfmac one is synchronous.
>>
>> Well, that depends on the eye of the beholder I guess. From user-space
>> perspective it is asynchronous regardless. A write access to the coredump
>> sysfs file eventually results in a uevent when the devcoredump entry is
>> created, ie. after driver has made a dev_coredump API call. Whether the
>> driver does that synchronously or asynchronously is irrelevant as far as
>> user-space is concerned.
>
> Is it really? The driver infrastructure seems to guarantee that the
> entirety of a driver's ->coredump() will complete before returning from
> the write. So it might be reasonable for some user to assume (based on
> implementation details, e.g., of brcmfmac) that the devcoredump will be
> ready by the time the write() syscall returns, absent documentation that
> says otherwise. But then, that's not how mwifiex works right now, so
> they might be surprised if they switch drivers.

Ok. I already agreed that the uevent behavior should be documented. The 
error handling you are bringing up below was something I realized as 
well. I am not familiar with mwifiex to determine what it can say about 
the coredump succeeding before scheduling the work.

> Anyway, *I'm* already personally used to these dumps being asynchronous,
> and writing tooling to listen for the uevent instead. But that doesn't
> mean everyone will be.
>
> Also, due to the differences in async/sync, mwifiex doesn't really
> provide you much chance for error handling, because most errors would be
> asynchronous. So brcmfmac's "coredump" has more chance for user programs
> to error-check than mwifiex's (due to the asynchronous nature) [1].
>
> BTW, I push on this mostly because this is migrating from a debugfs
> feature (that is easy to hand-wave off as not really providing a
> consistent/stable API, etc., etc.) to a documented sysfs feature. If it
> were left to rot in debugfs, I probably wouldn't be as bothered ;)

I appreciate it. The documentation is not in the stable ABI folder yet 
so I welcome any and all improvements. Might learn a thing or two from it.

>>> Brian
>>>
>>> [1] In fact, the ABI documentation really just describes kernel
>>> internals, rather than documenting any user-facing details, from what I
>>> can tell.
>>
>> You are right. Clearly I did not reach the end my learning curve here. I
>> assumed referring to the existing dev_coredump facility was sufficient, but
>> maybe it is worth a patch to be more explicit and mention the uevent
>> behavior. Also dev_coredump facility may be disabled upon which the trigger
>> will have no effect in sysfs. In the kernel the data passed by the driver is
>> simply freed by dev_coredump facility.
>
> Is there any other documentation for the coredump feature? I don't
> really see much.

Any other than the code itself you mean? I am not sure. Maybe Johannes 
knows.

> Brian
>
> [1] Oh wait, but I see that while ->coredump() has an integer return
> code...the caller ignores it:
>
> static ssize_t coredump_store(struct device *dev, struct device_attribute *attr,
> 			    const char *buf, size_t count)
> {
> 	device_lock(dev);
> 	if (dev->driver->coredump)
> 		dev->driver->coredump(dev);
> 	device_unlock(dev);
>
> 	return count;
> }
> static DEVICE_ATTR_WO(coredump);
>
> Is that a bug or a feature?

Yeah. Let's call it a bug. Just not sure what to go for. Return the 
error or change coredump callback to void return type.

Regards,
Arend