Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1650006 > unrolled thread

Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-05-25 00:40 +0200
    Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback Andy Lutomirski <luto@kernel.org> - 2017-05-25 06:20 +0200
      Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Fuzzey, Martin" <mfuzzey@parkeon.com> - 2017-05-25 10:30 +0200
        Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback ebiederm@xmission.com (Eric W. Biederman) - 2017-05-26 13:20 +0200
          Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-05-26 21:50 +0200
            Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback Dmitry Torokhov <dmitry.torokhov@gmail.com> - 2017-05-27 03:30 +0200
              Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Dmitry Torokhov <dmitry.torokhov@gmail.com> - 2017-05-27 03:30 +0200
                Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-05 22:30 +0200
                  Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Martin Fuzzey <mfuzzey@parkeon.com> - 2017-06-06 11:10 +0200
                    Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-06 18:40 +0200
                      Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-06 20:00 +0200
                  Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Alan Cox <alan@linux.intel.com> - 2017-06-06 17:00 +0200
                    Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-06 18:50 +0200
                      Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-06 20:00 +0200
                      Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Theodore Ts'o <tytso@mit.edu> - 2017-06-07 00:20 +0200
                        Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-07 02:30 +0200
                          Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback Andy Lutomirski <luto@kernel.org> - 2017-06-07 07:00 +0200
                            Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Dmitry Torokhov <dmitry.torokhov@gmail.com> - 2017-06-07 08:30 +0200
                              Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Alan Cox <alan@linux.intel.com> - 2017-06-07 14:30 +0200
                                Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-07 19:20 +0200
                              Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback Andy Lutomirski <luto@kernel.org> - 2017-06-09 03:20 +0200
                                Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-09 03:40 +0200
                                  Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-09 23:30 +0200
              Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-05-27 03:30 +0200
        Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-05-26 21:50 +0200
          Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-05-27 03:30 +0200
          Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Fuzzey, Martin" <mfuzzey@parkeon.com> - 2017-05-27 03:50 +0200
        Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-07 19:10 +0200
          Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Martin Fuzzey <mfuzzey@parkeon.com> - 2017-06-07 20:00 +0200
            Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-09 03:20 +0200
              Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-09 04:00 +0200
                Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback Martin Fuzzey <mfuzzey@parkeon.com> - 2017-06-09 09:50 +0200
                  Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-09 23:20 +0200
                  Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on  fallback "Luis R. Rodriguez" <mcgrof@kernel.org> - 2017-06-10 01:00 +0200

Page 2 of 2 — ← Prev page 1 [2]

#1661849

On Tue, Jun 6, 2017 at 11:25 PM, Dmitry Torokhov
<dmitry.torokhov@gmail.com> wrote:
> On Tue, Jun 06, 2017 at 09:56:47PM -0700, Andy Lutomirski wrote:
>> On Tue, Jun 6, 2017 at 5:22 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
>> > On Tue, Jun 06, 2017 at 06:11:51PM -0400, Theodore Ts'o wrote:
>> >> On Tue, Jun 06, 2017 at 06:47:34PM +0200, Luis R. Rodriguez wrote:
>> >> > On Tue, Jun 06, 2017 at 03:53:16PM +0100, Alan Cox wrote:
>> >
>> > We rely on swait, and swait right now only uses -ERESTARTSYS. Are
>> > you saying we could mask out -ERESTARTSYS and map it to -ERESTARTNOINTR
>> > or -ERESTARTNOHAND if we see fit for some future functionality / need ?
>>
>> I think that has essentially nothing to do with swait.  User code does
>> some syscall.  That syscall triggers a firmware load.  The caller gets
>> a signal.  If you're going to let firmware load get interrupted, you
>> need to consider what the syscall is.
>
> I think it is way too complicated and I do not think driver writers will
> stand a chance of implementing this correctly, given that often firmware
> load might be triggered indirectly and by multitude of syscalls.
>

That's what I meant, but I said it unclearly.  I meant that, if we're
going to start allowing interruption, we would need to audit all the
callers.  Ugh.

I suppose we could have request_firmware_interruptable(), but that
seems like it's barely worth it.

--Andy

[toc] | [prev] | [next] | [standalone]

#1661853

On Thu, Jun 8, 2017 at 6:14 PM, Andy Lutomirski <luto@kernel.org> wrote:
> On Tue, Jun 6, 2017 at 11:25 PM, Dmitry Torokhov
> <dmitry.torokhov@gmail.com> wrote:
>> On Tue, Jun 06, 2017 at 09:56:47PM -0700, Andy Lutomirski wrote:
>>> On Tue, Jun 6, 2017 at 5:22 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
>>> > On Tue, Jun 06, 2017 at 06:11:51PM -0400, Theodore Ts'o wrote:
>>> >> On Tue, Jun 06, 2017 at 06:47:34PM +0200, Luis R. Rodriguez wrote:
>>> >> > On Tue, Jun 06, 2017 at 03:53:16PM +0100, Alan Cox wrote:
>>> >
>>> > We rely on swait, and swait right now only uses -ERESTARTSYS. Are
>>> > you saying we could mask out -ERESTARTSYS and map it to -ERESTARTNOINTR
>>> > or -ERESTARTNOHAND if we see fit for some future functionality / need ?
>>>
>>> I think that has essentially nothing to do with swait.  User code does
>>> some syscall.  That syscall triggers a firmware load.  The caller gets
>>> a signal.  If you're going to let firmware load get interrupted, you
>>> need to consider what the syscall is.
>>
>> I think it is way too complicated and I do not think driver writers will
>> stand a chance of implementing this correctly, given that often firmware
>> load might be triggered indirectly and by multitude of syscalls.
>>
>
> That's what I meant, but I said it unclearly.  I meant that, if we're
> going to start allowing interruption, we would need to audit all the
> callers.  Ugh.

There are actually two audits worth evaluating if what we've concluded
is fair game:

  a) firmware sync calls on interruptible paths
  b) use of swait / old interruptible waits on sysfs paths

As for a) we drove tons of code away from using sync, request
firmware, and on async firmware the return value is lost, we only can
currently know if a failure of some sort occurred. If that push to
async failed we also scared away tons of drivers to use
request_firmware() calls on init and later incorrectly on probe due to
serialized init + probe delay on boot. From what I recall my last
Coccinelle evil-monster-hunt, I did indeed find quite a bit of drivers
still still relying on sync firmware which ultimately revealed use on
a probe path. The signal however was only effective as of commit
0cb64249ca500 ("firmware_loader: abort request if wait_for_completion
is interrupted") merged as of v4.0. Creating awareness of the issue
seems fair, but I don't think its worth a huge fly swatter.

I have no information to contribute for b) other than I was reluctant
to even consider it.

> I suppose we could have request_firmware_interruptable(), but that
> seems like it's barely worth it.

I don't think that's worth it, given the signal was effective only as
of v4.0, we already had a big push away from sync requests, and also
had the "don't use request firmware" on init scare which also
propagated to probe later.

  Luis

[toc] | [prev] | [next] | [standalone]

#1662759

On Thu, Jun 8, 2017 at 6:33 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> On Thu, Jun 8, 2017 at 6:14 PM, Andy Lutomirski <luto@kernel.org> wrote:
>> That's what I meant, but I said it unclearly.  I meant that, if we're
>> going to start allowing interruption, we would need to audit all the
>> callers.  Ugh.
>
> There are actually two audits worth evaluating if what we've concluded
> is fair game:
>
>   a) firmware sync calls on interruptible paths
>   b) use of swait / old interruptible waits on sysfs paths

And as I noted in the other thread -- another possible issue could be
any swait / interruptable wait on init or probe. Provided any child
completes and the kernel code for wait handler does abort, that
request would be terminated. This could for instance happen at bootup
as modules load and any child from the loader terminates.

We already have Coccinelle grammar to hunt for "though shall not
request firmware on init or probe", such SmPL grammar could be in turn
be repruposed to hunt for these types of conditions.

  Luis

[toc] | [prev] | [next] | [standalone]

#1651687

On Fri, May 26, 2017 at 2:26 PM, Dmitry Torokhov
<dmitry.torokhov@gmail.com> wrote:
> On Fri, May 26, 2017 at 12:46 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
>> On Fri, May 26, 2017 at 06:09:29AM -0500, Eric W. Biederman wrote:
>>> "Fuzzey, Martin" <mfuzzey@parkeon.com> writes:
>>> >>>> Maybe SIGCHLD shouldn't interrupt firmware loading?
>>> >
>>> > I don't think there's a way of doing that without disabling all
>>> > signals (ie using the non interruptible wait variants).
>>> > It used to be that way (which is why I only ran into this after
>>> > updating from an ancient 3.16 kernel to a slightly less ancient 4.4)
>>> > But there are valid reasons for wanting to be able to interrupt
>>> > firmware loading (like being able to kill the userspace helper)
>>>
>>> Perhaps simply using a killable wait and not a fully interruptible
>>> wait would be better?
>>
>> What do you mean by a killable wait BTW?
>
> https://lwn.net/Articles/288056/
>
> I think only interrupting firmware loading with fatal signals would
> make a lot of sense.
>
>>
>> ret = swait_event_interruptible_timeout() is being used right now.
>
> It looks like we are missing swait_event_killable*(), but I do not
> think it would be hard to add.

What should we do for stable ? Is this a *stable* issue ?

 Luis

[toc] | [prev] | [next] | [standalone]

#1651592 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On Thu, May 25, 2017 at 10:28:38AM +0200, Fuzzey, Martin wrote:
> On 25 May 2017 at 06:13, Andy Lutomirski <luto@kernel.org> wrote:
> >>>
> >>> Can you give a simple example of what's going on and why it matters?
> >>>
> 
> 
> Here is the use case in which I ran into this problem.
> 
> I have a driver which does request_firmware() when a write() is done
> to a sysfs file.
> 
> The write() was being done by an android init script (with the init
> interpreter "write" command).
> init, of course, forks lots of processes and some of the children die.
> 
> So the scenario was the following:
> 
> 1) Android init calls write() on the sysfs file
> 2) The sysfs .store() callback registered by a driver is called
> 3) The driver calls request_firmware()
> 4) request_firmware() sends the firmware load request to userspace and
> calls wait_for_completion_interruptible()

No no, this is not how the fallback system works.

Please have a look at the latest documentation revamp [0], in particular
the fallback mechanism [1], hopefully this clarifies other details you might
like to know about this first part.

Please note I *highly* encourage relying on the uevent fallback mechanism,
given the custom fallback mechanism has its own quirks and from my own review
I'd be pretty wary about using it.

[0] https://www.kernel.org/doc/html/latest/driver-api/firmware/index.html
[1] https://www.kernel.org/doc/html/latest/driver-api/firmware/fallback-mechanisms.html

Here is the correct summary:

1) Driver issues request_firmware()
2) firmware_class looks for the firmware on the local filesystem, if it
   does not find it and the fallback mechanism is enabled
   firmware_class issues uevent for the device that requested the firmware
3) Userspace is expected to be monitoring for these uevents and based on
   them should only *then* try a write() on the respective sysfs file.
4) The kernel actually uses swait_event_interruptible_timeout() as of
   v4.10 due to commit 5b029624948d6 ("firmware: do not use fw_lock for
   fw_state protection"), before it used wait_for_completion_interruptible_timeout().
   It uses this to *wait* for the sysfs write to complete, so can be
   woken up by it.

Please note we just uncovered another issue on firmware_class reported by
Nicolas when multiple cards are used and a signal of completion on sucesss or
error is not issued [2] for which John Ewalt has suggested a fix for [3].
I just reviewed this yesterday and determiend the swake_up()/swake_up_all()
fix is one atomic fix for 5b029624948d6 which moved us to swait given otherwise
we won't wake up all waiters on the same file. The wait stuff is used for
other means not well documented than the sysfs interface: if multiple 
requests come in for the same file we batch up these requests as one and
wait for a signal so we can rely on the same file fetch to write to the
pending request. Unfortunately I have discovered yesterday we have quite a
bit of fail paths that never sent the swake_up_all() (this would be
a completion on kernels older than v4.10) on *many* error paths. Note
though there is only one wait used and it uses:

	swait_event_interruptible_timeout()	>= v4.10
	wait_for_completion_interruptible_timeout() older

Since this is interruptible, and has a timeout it does mean if we have
a reboot the wait will be killed.

Since this wait also has a timeout it means we will have an end if
userspace never does anything.

5) Two sysfs files are exposed:

        /sys/$DEVPATH/loading
		echo 1 > loading # starts load
		echo 0 > loading # indicates load is done
				 # asking to load data to driver
		echo -1 > loading # indicates an error occured

For detail see firmware_loading_store(), note we check for fw_state_is_aborted()
before proceeding.

        /sys/$DEVPATH/data

For details see firmware_data_read(), note it checks for fw_state_is_done()
early on before proceeding. fw_state_is_done() includes an abort check:

static inline bool __fw_state_is_done(enum fw_status status)                    
{                                                                               
        return status == FW_STATUS_DONE || status == FW_STATUS_ABORTED;         
}   

[2] https://bugzilla.kernel.org/show_bug.cgi?id=195477
[3] https://bugzilla.kernel.org/attachment.cgi?id=256493&action=diff&collapsed=&headers=1&format=raw

> 5) A child dies and raises SIGCHLD

What child? The process doing the write() ?

If we have userspace doing a write() on the sysfs interface then
firmware_loading_store() gets kicked off, it should actually try to complete,
unless the singal was processed early in which case the write would bail early
on fw_state_is_done() check.

Otherwise I suppose the kernel can send a signal pending to the swait but
its not clear under what conditions exactly.

> 6) wait_for_completion_interruptible() returns -ERESTARTSYS due to the signal

That's for kernels older than v4.10, on >= v4.10 that would be
swait_event_interruptible_timeout() and I confirm -ERESTARTSYS can
be sent as a return value.

> 7) request_firmware() [before this patch] translated that to -EAGAIN

Right, this is today's default error if we fail on the wait. This error would
be returned to the *driver*. This is returned on _request_firmware_load() and
sent back to the request_firmware_*() call.

Martin seems to be arguing -ERESTARTSYS should be sent back instead given
it is what the wait returned after all.

> 8) The driver (in my case) ignored this [because the firmware was not
> critical - it was for checking if a microcontroler was up to date]
> (but it could have returned it to userspace, same problem)

I agree that using -EAGAIN for a signal is incorrect and we should propagate
something more reasonable to the driver, but also given that firmware loading
is typically a bad idea on init but is acceptable on probe, it could mean a
driver propagates this to userspace.

Another issue to consider here is drivers using request_firmware_nowait() may
not get an actual error passed down given the lack of semantics of the
firmware_class. This will be fixed later on the driver data API but for now
it means async callbacks can only infer what might have happened. The callback
must be revised in this light to ensure some sanity check is done of some sort.

So to say that the driver ignored it is saying that the sync call was used
for sure and we know that it ignored -EAGAIN. Is that the case ?

If not what driver was used that ignored -EAGAIN? If it was an async callback
no error value is passed and what you are dealing with is more of an issue
with lack of semantics in the callback very likely and the inability to
deal with error respectively.

So we have three separate issues here:

1) A driver's possible lack of propagation errors
2) The actual error returned by firmware_class and semantics on async callback
3) What errors we should send on syfs files if a file if interrupted or error

The gruseome TLDR details below.

=========================================================

1) A driver's possible lack of propagation errors

Given 2) drivers should take care to be pedantic about how they deal with
errors. In particular if an async callback is used, a signal received
may mean the callback can only check for NULL, and it can only infer if
an error occurred. If the driver deals with daisy chaining files it might
mean it could skip over files without treating fatal errors accordingly.

2) The actual error returned by firmware_class and semantics on async callback

The async callbacks only get NULL on the firmware passed if an error was
detected, it cannot know if a specific error actually occurred. Currently only
sync calls can get the actual -EAGAIN (and if we stich this the respective
-ERESTARTSYS). This will be fixed with the driver data API, as it allows
callbacks to get the actual error passed.

Also, since async callbacks can be used on probe, but we don't want to delay
boot and since init + probe() are batched together we tend to not want to wait
on probe, so this is why many drivers returns right away after an async
firmware call. Even though we have async probe now (and drivers can prefer
this) that still would mean userspace does not get the actual error value. Its
up to drivers to also be able to take down the driver on error if an async call
for fw is used and it fails. The actual error on an async call will only be
available in the future, right now only NULL is passed.

Care with the above lack of current semantics must be taken into account

A user API question still stands as to if we want to send -ERESTARTSYS to
drivers to potentially send back to userspace on driver load given this
can still be sent for sync calls today and we are sending -EAGAIN.

3) What errors we should send on syfs files if a file if interrupted or error

We could capture an error happened on any of the sysfs files by checking
the firwmware status. For instance fw_state_is_aborted() checks on
firmware_loading_store(); fw_state_is_done() on firmware_data_read();
or fw_state_is_done() check on firmware_data_write().

Right now the errors we pass vary. As an example firmware_data_write()
can fail with -ENODEV if for some reason a signal was sent and we
bailed early. That -EPERM on !capable(CAP_SYS_RAWIO) should probably
be changed to -EACCESS. The firmware_data_read() sends -ENODEV() on
error. firmware_loading_store() sends the size of the buffer passed
to us on error -- that seems like an issue.

> The point being that, due to a signal (SIGCHLD) which has nothing to
> do with the firmware loading process, the firmware load was not done.

Who gets SIGCHLD and how does this exactly get to swait_event_interruptible_timeout()
or wait_for_completion_interruptible_timeout() exactly ?

> Also EAGAIN is the same error used if the load request times out so it
> was impossible to distinguish the two cases.

I agree this is an issue.

> ERESTARTSYS is an internal error and is not returned to userspace.
> Instead it is handled by the linux syscall machinery which,

Not accurate, we just never pass it on the firmware_class.c on
swait_event_interruptible_timeout() error. In fact we mask all
errors to -EAGAIN, and if the timeout ran out we just send
-ETIMEDOUT. If the request was aborted early the error -ENOENT
is sent.

> after
> processing the signal either restarts (transpently to userspace) the
> syscall or returns EINTR to userspace (depending if the signal handler
> users SA_RESTART - see man 7 signal)
> 
> 
> With this patch here is what happens:
> 
> 1) Android init calls write() on the sysfs file
> 2) The sysfs .store() callback registered by a driver is called
> 3) The driver calls request_firmware()
> 4) request_firmware() sends the firmware load request to userspace and
> calls wait_for_completion_interruptible()
> 5) A child dies and raises SIGCHLD
> 6) wait_for_completion_interruptible() returns -ERESTARTSYS due to the signal
> 7) request_firmware() [with this patch] returns -ERESTARTSYS
> 8) The driver returns -ERSTARTSYS from its sysfs .store method

If you're talking about a custom driver of sorts that triggered
a request_firmware() call (note sync) then yes your earlier description
is accurate and in this case as well the driver specific sysfs interface
can end up returning the same error.

If this file is custom then its up to you to decide what you want
for error on that file, but for the firmware_class.c I do agree on
returning an agreed upon error so drivers can Do The Right Thing (TM).

> 9) The system call machinery invokes the signal handler
> 10) The signal handler does its stuff
> 11) Because SA_RESTART was set the system call is restarted (calling
> the sysfs .store) and we try it all again from step 2

OK so this seems to reveal an internal working of some android
loader of some sort. Thanks.

> Note that, on the the userspace side  write() is only called once

The write is to the custom driver trigger which calls request_firmware() ?

> (the
> restart is transparent to userspace which is oblivious to all this)
> The kernel side write() (which calls .store() is called multiple times
> (so that code does need to know about this)

Even if the signal is captured and -ERSTARTSYS is sent down to a custom
sysfs file trigger which called request_firmware(), if -ERSTARTSYS
triggers userspace to try again the old sysfs loader for the fallback
mechanism will not be available anymore, so actually I'd be surprised
if the retry designed by android works here. It *would* have to
retry the whole thing again, starting from the custom trigger.

> >>> ERESTARTSYS and friends are highly magical, and I'm not convinced that
> >>> allowing _request_firmware_load to return -ERESTARTSYS is actually a
> >>> good idea.  What if there are system calls that can't handle this
> >>> style of restart that start being restarted as a result?

Given the above explanation I see the concern now. In our case I believe a
valid concern would be sending -ERESTARTSYS to the system call finit_module()
given request_firmware() could be called on probe and if a stupid driver did a
sync call on probe it could potentially send -ERESTARTSYS down.

This could mean we get a loop on finit_module() if a signal is detected
on firmware_request() on every call. Is that fine? Is it expected ?

> If the caller is unable to restart (for example if the driver's
> .store() callback had already done lots of stuff that couldn't be
> undone) it is free to translate -ERSTARTSYS to -EINTR before
> returning.
> But request_frimware() can't know about that.

So you seem to be suggesting the driver's should decide to mask or
not -ERSTARTSYS.

Do we send -ERSTARTSYS on any other paths on finit_module() ?

> >>> Maybe SIGCHLD shouldn't interrupt firmware loading?
> 
> I don't think there's a way of doing that without disabling all
> signals (ie using the non interruptible wait variants).

It could be an option if we deem such API is needed.

> It used to be that way (which is why I only ran into this after
> updating from an ancient 3.16 kernel to a slightly less ancient 4.4)

Indeed and back then I think we only ever returned -ENOMEM on error !

> But there are valid reasons for wanting to be able to interrupt
> firmware loading (like being able to kill the userspace helper)

Let's call it the fallback mechanism, as we should. OK so if the
fallback helper is called I fail to see how this is detrimental,
we'd just timeout, no ?

  Luis

[toc] | [prev] | [next] | [standalone]

#1651622 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On Fri, May 26, 2017 at 10:23:03PM +0200, Fuzzey, Martin wrote:
> On 26 May 2017 at 21:40, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> >
> > No no, this is not how the fallback system works.
> 
> The sysfs file I was talking about is *not* the sysfs file involved in
> the firmware loading mechanism
> but a *custom* sysfs file used by a driver to *trigger* the call to
> request_firmware() [synchronous] in the first place.

OK!

> That is, my driver does not do request_firmware() in its probe()
> function like most but when requested by
> userspace. That's a valid usage as far as I can tell. Nothing says
> request_firmware() is only allowed from probe.

In fact we prefer sync calls not to be done on probe from just a practical
perspective as this can delay boot.

> Not sure it is relevant here but here's the reason for doing it this
> way: (skip this bit if not interested)
> I have a small microcontroller used to manage the power to the main CPU.
> It is connected to the CPU by a GPIO line.
> On power up the microcrontroller powers up the main CPU and starts a timer.
> If the application doesn't start in time the microcontroller power
> cycles the CPU.

OK

> In addition to the the above the microcontroller is connected to the
> CPU by a I2C bus for various other functions
> including firmware update (of the microcontroller software).

OK.

> In order to keep the critical power up code on the microcontroller
> very simple the I2C connection is not available
> until *after* the power up confirmation by the GPIO line.

I was not able to parse this.

You seem to be saying the I2C interface for firmware updates of the
microcontroller software is not available until the CPU it controls
is up, and this is confirmed through the GPIO line ? The I2c firmware
update interface is not for the target CPU the microcontroller controls ?

From reading the below it seems this is right.

> So, on the linux side, there is a custom driver which exports a sysfs
> entry that userspace writes to in order to confirm
> startup.

OK so Linux runs on this host the microcontroller powers up, and it has
a custom driver, and it just toggles a knob to inform the microcontroller,
"Hey dude, I'm up, give me firmware!" ?

> The driver, when that sysfs file is written to, first sets the GPIO
> line to signal the microcontroller that the application has
> started so it can stop its timer and activate the I2C interface.

Got it! How does it know the I2C interface is ready ? Does it wait ?

> Then it does a request_firmware() to obtain the firmware the microcontroler
> is supposed to have. It then verifies it using commands over I2C to compare
> checksums etc and updates it if needed.

OK. Got it.

> >> 9) The system call machinery invokes the signal handler
> >> 10) The signal handler does its stuff
> >> 11) Because SA_RESTART was set the system call is restarted (calling
> >> the sysfs .store) and we try it all again from step 2
> >
> > OK so this seems to reveal an internal working of some android
> > loader of some sort. Thanks.
> >
> 
> Nothing Android specific here.
> 
> This is the standard *linux* behaviour.
> The retry is done *in the kernel* not on the android userspace side

Alright !

> > This could mean we get a loop on finit_module() if a signal is detected
> > on firmware_request() on every call. Is that fine? Is it expected ?
> 
> Ok so here we are talking about the standard case 

I would not say its standard. We support it, but we prefer async requests
are used there.

But its possible...  And since this as you clarify Linux deals with this
specially it is of interest what we do return.

> of request_firmware() being
> called from probe() If the driver is a loadable module then yes the it will
> be retried.  If it gets a signal on every try then something else is wrong
> I'd say
> 
> If the driver is compiled in then there is no retry since the call to probe
> doesn't go through the syscall machinary. But there shouldn't be a
> signal either in that
> case since it's not being called from a userspace process in that case.

Still -- its all possible on finit_module() so we still must consider the case.
If sending -ERESTARTSYS is not a good idea we could for instance send another
special error to let driver *then* make an informed decision, but that it
could still interpret as it actually having been -ERESTARTSYS.

> > So you seem to be suggesting the driver's should decide to mask or
> > not -ERSTARTSYS.
> >
> 
> Only if the driver knows that it is not restartable *itself*.
> Shoudn't happen very often

I'd prefer to hear more from linux-api folks.

  Luis

[toc] | [prev] | [next] | [standalone]

#1651716

On 26 May 2017 at 21:40, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> On Thu, May 25, 2017 at 10:28:38AM +0200, Fuzzey, Martin wrote:
>> On 25 May 2017 at 06:13, Andy Lutomirski <luto@kernel.org> wrote:
>> >>>
>> >>> Can you give a simple example of what's going on and why it matters?
>> >>>
>>
>>
>> Here is the use case in which I ran into this problem.
>>
>> I have a driver which does request_firmware() when a write() is done
>> to a sysfs file.
>>
>> The write() was being done by an android init script (with the init
>> interpreter "write" command).
>> init, of course, forks lots of processes and some of the children die.
>>
>> So the scenario was the following:
>>
>
> No no, this is not how the fallback system works.
>

I was (implicitly) describing the fallback mechanism.
Indeed as you say the kernel now first tries to load the firmware
itself without involving userspace.
I'm describing what happens *after* this fails and the kernel falls
back to userspace

The sysfs file I was talking about is *not* the sysfs file involved in
the firmware loading mechanism
but a *custom* sysfs file used by a driver to *trigger* the call to
request_firmware() [synchronous] in the first place.

That is, my driver does not do request_firmware() in its probe()
function like most but when requested by
userspace. That's a valid usage as far as I can tell. Nothing says
request_firmware() is only allowed from probe.

Not sure it is relevant here but here's the reason for doing it this
way: (skip this bit if not interested)
I have a small microcontroller used to manage the power to the main CPU.
It is connected to the CPU by a GPIO line.
On power up the microcrontroller powers up the main CPU and starts a timer.
If the application doesn't start in time the microcontroller power
cycles the CPU.

In addition to the the above the microcontroller is connected to the
CPU by a I2C bus for various other functions
including firmware update (of the microcontroller software).
In order to keep the critical power up code on the microcontroller
very simple the I2C connection is not available
until *after* the power up confirmation by the GPIO line.

So, on the linux side, there is a custom driver which exports a sysfs
entry that userspace writes to in order to confirm
startup.
The driver, when that sysfs file is written to, first sets the GPIO
line to signal the microcontroller that the application has
started so it can stop its timer and activate the I2C interface. Then
it does a request_firmware() to obtain the firmware
the microcontroler is supposed to have. It then verifies it using
commands over I2C to compare checksums etc and
updates it if needed.


>> 5) A child dies and raises SIGCHLD
>
> What child? The process doing the write() ?
>
A child of the process doing the write on the drivers custom sysfs
entry that triggered the request_firmware()

> Martin seems to be arguing -ERESTARTSYS should be sent back instead given
> it is what the wait returned after all.
>

Yes


> If you're talking about a custom driver of sorts that triggered
> a request_firmware() call (note sync) then yes your earlier description
> is accurate and in this case as well the driver specific sysfs interface
> can end up returning the same error.
>

Exactly


> If this file is custom then its up to you to decide what you want
> for error on that file, but for the firmware_class.c I do agree on
> returning an agreed upon error so drivers can Do The Right Thing (TM).
>

Yes


>> 9) The system call machinery invokes the signal handler
>> 10) The signal handler does its stuff
>> 11) Because SA_RESTART was set the system call is restarted (calling
>> the sysfs .store) and we try it all again from step 2
>
> OK so this seems to reveal an internal working of some android
> loader of some sort. Thanks.
>

Nothing Android specific here.

This is the standard *linux* behaviour.
The retry is done *in the kernel* not on the android userspace side


>> Note that, on the the userspace side  write() is only called once
>
> The write is to the custom driver trigger which calls request_firmware() ?
>

Yes



> This could mean we get a loop on finit_module() if a signal is detected
> on firmware_request() on every call. Is that fine? Is it expected ?

Ok so here we are talking about the standard case of
request_firmware() being called from probe()
If the driver is a loadable module then yes the it will be retried.
If it gets a signal on every try then something else is wrong I'd say

If the driver is compiled in then there is no retry since the call to probe
doesn't go through the syscall machinary. But there shouldn't be a
signal either in that
case since it's not being called from a userspace process in that case.

> So you seem to be suggesting the driver's should decide to mask or
> not -ERSTARTSYS.
>

Only if the driver knows that it is not restartable *itself*.
Shoudn't happen very often


Martin

[toc] | [prev] | [next] | [standalone]

#1660029 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On Thu, May 25, 2017 at 10:28:38AM +0200, Fuzzey, Martin wrote:
> 1) Android init calls write() on the sysfs file
> 2) The sysfs .store() callback registered by a driver is called
> 3) The driver calls request_firmware()
> 4) request_firmware() sends the firmware load request to userspace and
> calls wait_for_completion_interruptible()

Martin, just for completeness on documenting on the commit log of the next
swait proposed fix for this -- what signal did the process get from which you
note the child dies below ? Exactly what in Android sent this signal ?

> 5) A child dies and raises SIGCHLD
> 6) wait_for_completion_interruptible() returns -ERESTARTSYS due to the signal

  Luis

[toc] | [prev] | [next] | [standalone]

#1660053 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On 07/06/17 19:08, Luis R. Rodriguez wrote:
> On Thu, May 25, 2017 at 10:28:38AM +0200, Fuzzey, Martin wrote:
>> 1) Android init calls write() on the sysfs file
>> 2) The sysfs .store() callback registered by a driver is called
>> 3) The driver calls request_firmware()
>> 4) request_firmware() sends the firmware load request to userspace and
>> calls wait_for_completion_interruptible()
> Martin, just for completeness on documenting on the commit log of the next
> swait proposed fix for this -- what signal did the process get from which you
> note the child dies below ? Exactly what in Android sent this signal ?

Android didn't send the signal, the kernel did (SIGCHLD).

Like this:

1) Android init (pid=1) fork()s (say pid=42) [this child process is 
totally unrelated to firmware loading]
2) Android init (pid=1) does a write() on a (driver custom) sysfs file 
which ends up calling request_firmware() kernel side
3) The firmware loading fallback mechanism is used, the request is sent 
to userspace and pid 1 waits in the kernel on wait_*
4) before firmware loading completes pid 42 dies (for any reason - in my 
case normal termination)
5) Kernel delivers SIGCHLD to pid=1 to tell it a child has died, which 
causes -ERESTARTSYS to be returned from wait_*


Martin

[toc] | [prev] | [next] | [standalone]

#1661851

On Wed, Jun 7, 2017 at 10:54 AM, Martin Fuzzey <mfuzzey@parkeon.com> wrote:
> On 07/06/17 19:08, Luis R. Rodriguez wrote:
>>
>> On Thu, May 25, 2017 at 10:28:38AM +0200, Fuzzey, Martin wrote:
>>>
>>> 1) Android init calls write() on the sysfs file
>>> 2) The sysfs .store() callback registered by a driver is called
>>> 3) The driver calls request_firmware()
>>> 4) request_firmware() sends the firmware load request to userspace and
>>> calls wait_for_completion_interruptible()
>>
>> Martin, just for completeness on documenting on the commit log of the next
>> swait proposed fix for this -- what signal did the process get from which
>> you
>> note the child dies below ? Exactly what in Android sent this signal ?
>
>
> Android didn't send the signal, the kernel did (SIGCHLD).
>
> Like this:
>
> 1) Android init (pid=1) fork()s (say pid=42) [this child process is totally
> unrelated to firmware loading]
> 2) Android init (pid=1) does a write() on a (driver custom) sysfs file which
> ends up calling request_firmware() kernel side
> 3) The firmware loading fallback mechanism is used, the request is sent to
> userspace and pid 1 waits in the kernel on wait_*
> 4) before firmware loading completes pid 42 dies (for any reason - in my
> case normal termination)

Interesting, could one interpretation here be that the process
successfully finishing + the signal being sent beats out the timing of
the firmware_class syfs code detecting that the write completed ?

> 5) Kernel delivers SIGCHLD to pid=1 to tell it a child has died, which
> causes -ERESTARTSYS to be returned from wait_*

  Luis

[toc] | [prev] | [next] | [standalone]

#1661860

On Thu, Jun 8, 2017 at 6:10 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> On Wed, Jun 7, 2017 at 10:54 AM, Martin Fuzzey <mfuzzey@parkeon.com> wrote:
>> On 07/06/17 19:08, Luis R. Rodriguez wrote:
>>>
>>> On Thu, May 25, 2017 at 10:28:38AM +0200, Fuzzey, Martin wrote:
>>>>
>>>> 1) Android init calls write() on the sysfs file
>>>> 2) The sysfs .store() callback registered by a driver is called
>>>> 3) The driver calls request_firmware()
>>>> 4) request_firmware() sends the firmware load request to userspace and
>>>> calls wait_for_completion_interruptible()
>>>
>>> Martin, just for completeness on documenting on the commit log of the next
>>> swait proposed fix for this -- what signal did the process get from which
>>> you
>>> note the child dies below ? Exactly what in Android sent this signal ?
>>
>>
>> Android didn't send the signal, the kernel did (SIGCHLD).
>>
>> Like this:
>>
>> 1) Android init (pid=1) fork()s (say pid=42) [this child process is totally
>> unrelated to firmware loading]
>> 2) Android init (pid=1) does a write() on a (driver custom) sysfs file which
>> ends up calling request_firmware() kernel side
>> 3) The firmware loading fallback mechanism is used, the request is sent to
>> userspace and pid 1 waits in the kernel on wait_*
>> 4) before firmware loading completes pid 42 dies (for any reason - in my
>> case normal termination)

Martin just to be clear, by "normal case termination" do you mean
completing successfully ?? Ie the firmware actually did make it onto
the device ?

> Interesting, could one interpretation here be that the process
> successfully finishing + the signal being sent beats out the timing of
> the firmware_class syfs code detecting that the write completed ?
>
>> 5) Kernel delivers SIGCHLD to pid=1 to tell it a child has died, which
>> causes -ERESTARTSYS to be returned from wait_*
>
>   Luis

[toc] | [prev] | [next] | [standalone]

#1662045 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On 09/06/17 03:57, Luis R. Rodriguez wrote:
> On Thu, Jun 8, 2017 at 6:10 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
>>> Android didn't send the signal, the kernel did (SIGCHLD).
>>>
>>> Like this:
>>>
>>> 1) Android init (pid=1) fork()s (say pid=42) [this child process is totally
>>> unrelated to firmware loading]
>>> 2) Android init (pid=1) does a write() on a (driver custom) sysfs file which
>>> ends up calling request_firmware() kernel side
>>> 3) The firmware loading fallback mechanism is used, the request is sent to
>>> userspace and pid 1 waits in the kernel on wait_*
>>> 4) before firmware loading completes pid 42 dies (for any reason - in my
>>> case normal termination)
> Martin just to be clear, by "normal case termination" do you mean
> completing successfully ?? Ie the firmware actually did make it onto
> the device ?

The firmware did *not* make it onto the device since the 
request_firmware() call returned an error
(the code that would have transfered it to the device is only executed 
following a successful request_firmware)

The process that terminates normally is unrelated to firmware loading as 
I said above.

The only things that matter are:
- It is a child process of the process that calls request_firmware()
- It terminates *while* the the wait_ is still in progress


Here is a way of reproducing the problem using the test_firmware module 
(which I only just saw) on normal linux with no Android or custom driver


#!/bin/sh
set -e

# Make sure the system firmware loader doesn't get in the way
/etc/init.d/udev stop

modprobe test_firmware

DIR=/sys/devices/virtual/misc/test_firmware

echo 10 >/sys/class/firmware/timeout;
sleep 2 &
echo -n "/some/non/existing/file.bin" > "$DIR"/trigger_request;



If run with the "sleep 2 &" it terminates after 2 seconds
If the sleep is commented it runs for the expected 10 seconds (the 
firmware loading timeout)

Since the sleep process is a child of the script process requesting a 
firmware load its death causes a SIGCHLD causing request_firmware() to 
abort prematurely.


Martin

[toc] | [prev] | [next] | [standalone]

#1662756 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On Fri, Jun 09, 2017 at 09:40:47AM +0200, Martin Fuzzey wrote:
> On 09/06/17 03:57, Luis R. Rodriguez wrote:
> > On Thu, Jun 8, 2017 at 6:10 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> > > > Android didn't send the signal, the kernel did (SIGCHLD).
> > > > 
> > > > Like this:
> > > > 
> > > > 1) Android init (pid=1) fork()s (say pid=42) [this child process is totally
> > > > unrelated to firmware loading]
> > > > 2) Android init (pid=1) does a write() on a (driver custom) sysfs file which
> > > > ends up calling request_firmware() kernel side
> > > > 3) The firmware loading fallback mechanism is used, the request is sent to
> > > > userspace and pid 1 waits in the kernel on wait_*
> > > > 4) before firmware loading completes pid 42 dies (for any reason - in my
> > > > case normal termination)
> > Martin just to be clear, by "normal case termination" do you mean
> > completing successfully ?? Ie the firmware actually did make it onto
> > the device ?
> 
> The firmware did *not* make it onto the device since the request_firmware()
> call returned an error
> (the code that would have transfered it to the device is only executed
> following a successful request_firmware)
> 
> The process that terminates normally is unrelated to firmware loading as I
> said above.
> 
> The only things that matter are:
> - It is a child process of the process that calls request_firmware()
> - It terminates *while* the the wait_ is still in progress
> 
> 
> Here is a way of reproducing the problem using the test_firmware module
> (which I only just saw) on normal linux with no Android or custom driver
> 
> 
> #!/bin/sh
> set -e
> 
> # Make sure the system firmware loader doesn't get in the way
> /etc/init.d/udev stop
> 
> modprobe test_firmware
> 
> DIR=/sys/devices/virtual/misc/test_firmware
> 
> echo 10 >/sys/class/firmware/timeout;
> sleep 2 &
> echo -n "/some/non/existing/file.bin" > "$DIR"/trigger_request;
> 
> 
> 
> If run with the "sleep 2 &" it terminates after 2 seconds
> If the sleep is commented it runs for the expected 10 seconds (the firmware
> loading timeout)
> 
> Since the sleep process is a child of the script process requesting a
> firmware load its death causes a SIGCHLD causing request_firmware() to abort
> prematurely.

Thanks this could mean we also *should* trigger a failure if init is issuing
modprobe on a series of drivers and one completes before another while
request_firmware() is called on init or probe on a subsequent driver. If true
I'm surprised this never was reported back when the fallback mechanism was
popular, I suppose it was not an issue given most firmware *was* present on
/lib/firmware/ and the direct filesystem lookup first step always found the
firmware first, so this would only be an issue for folks relying on the
fallback mechanism exclusively.

Will include a test case based on your above script. Thanks!

  Luis

[toc] | [prev] | [next] | [standalone]

#1662806 — Re: [PATCH v2] firmware: fix sending -ERESTARTSYS due to signal on fallback

On Fri, Jun 09, 2017 at 09:40:47AM +0200, Martin Fuzzey wrote:
> On 09/06/17 03:57, Luis R. Rodriguez wrote:
> > On Thu, Jun 8, 2017 at 6:10 PM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> > > > Android didn't send the signal, the kernel did (SIGCHLD).
> > > > 
> > > > Like this:
> > > > 
> > > > 1) Android init (pid=1) fork()s (say pid=42) [this child process is totally
> > > > unrelated to firmware loading]
> > > > 2) Android init (pid=1) does a write() on a (driver custom) sysfs file which
> > > > ends up calling request_firmware() kernel side
> > > > 3) The firmware loading fallback mechanism is used, the request is sent to
> > > > userspace and pid 1 waits in the kernel on wait_*
> > > > 4) before firmware loading completes pid 42 dies (for any reason - in my
> > > > case normal termination)
> > Martin just to be clear, by "normal case termination" do you mean
> > completing successfully ?? Ie the firmware actually did make it onto
> > the device ?
> 
> The firmware did *not* make it onto the device since the request_firmware()
> call returned an error
> (the code that would have transfered it to the device is only executed
> following a successful request_firmware)
> 
> The process that terminates normally is unrelated to firmware loading as I
> said above.
> 
> The only things that matter are:
> - It is a child process of the process that calls request_firmware()
> - It terminates *while* the the wait_ is still in progress
> 
> 
> Here is a way of reproducing the problem using the test_firmware module
> (which I only just saw) on normal linux with no Android or custom driver
> 
> 
> #!/bin/sh
> set -e
> 
> # Make sure the system firmware loader doesn't get in the way
> /etc/init.d/udev stop
> 
> modprobe test_firmware
> 
> DIR=/sys/devices/virtual/misc/test_firmware
> 
> echo 10 >/sys/class/firmware/timeout;
> sleep 2 &
> echo -n "/some/non/existing/file.bin" > "$DIR"/trigger_request;
> 
> 
> 
> If run with the "sleep 2 &" it terminates after 2 seconds
> If the sleep is commented it runs for the expected 10 seconds (the firmware
> loading timeout)
> 
> Since the sleep process is a child of the script process requesting a
> firmware load its death causes a SIGCHLD causing request_firmware() to abort
> prematurely.

BTW I've run some more tests and if you use request_firmware_nowait() the above
issue also is avoidable. Also, if you use the custom fallback (uevent parameter
is false) the timeout is max value always. The only way to kill the custom
requests is through the echo -1 > loading or the max timeout triggers (don't
rely on that!). The custom interface was a bad idea though so best to just
avoid it.

Anyway I can reproduce in my tests now, will add this testcase as well.
The stable fixes will be sent as well.

  Luis

[toc] | [prev] | [standalone]

Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | linux.kernel

csiph-web