RE: [PATCH] mmc: dw_mmc: Make sure we don't get stuck when we get an error

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sat, May 17, 2014, Doug Anderson wrote:
> Seungwon,
> 
> On Thu, May 15, 2014 at 6:46 PM, Seungwon Jeon <tgih.jun@xxxxxxxxxxx> wrote:
> > On Wed, May 14, 2014, Doug Anderson wrote:
> >> Seungwon,
> >>
> >> On Mon, May 12, 2014 at 9:52 PM, Seungwon Jeon <tgih.jun@xxxxxxxxxxx> wrote:
> >> > Hi Doug,
> >> >
> >> > On Tue, May 13, 2014, Doug Anderson wrote:
> >> >> Seungwon,
> >> >>
> >> >> On Sat, May 10, 2014 at 7:11 AM, Seungwon Jeon <tgih.jun@xxxxxxxxxxx> wrote:
> >> >> > On Fri, May 09, 2014, Sonny Rao wrote:
> >> >> >> On Thu, May 8, 2014 at 2:42 AM, Yuvaraj Kumar <yuvaraj.cd@xxxxxxxxx> wrote:
> >> >> >> > Any comments on this patch?
> >> >> >> >
> >> >> >>
> >> >> >> I'll just add that without this fix, running the tuning loop for UHS
> >> >> >> modes is not reliable on dw_mmc because errors will happen and you
> >> >> >> will eventually hit this race and hang.  This can happen any time
> >> >> >> there is tuning like during boot or during resume from suspend.
> >> >> >>
> >> >> >> > On Thu, Mar 27, 2014 at 11:48 AM, Yuvaraj Kumar C D
> >> >> >> > <yuvaraj.cd@xxxxxxxxx> wrote:
> >> >> >> >> From: Doug Anderson <dianders@xxxxxxxxxxxx>
> >> >> >> >>
> >> >> >> >> If we happened to get a data error at just the wrong time the dw_mmc
> >> >> >> >> driver could get into a state where it would never complete its
> >> >> >> >> request.  That would leave the caller just hanging there.
> >> >> >> >>
> >> >> >> >> We fix this two ways and both of the two fixes on their own appear to
> >> >> >> >> fix the problems we've seen:
> >> >> >> >>
> >> >> >> >> 1. Fix a race in the tasklet where the interrupt setting the data
> >> >> >> >>    error happens _just after_ we check for it, then we get a
> >> >> >> >>    EVENT_XFER_COMPLETE.  We fix this by repeating a bit of code.
> >> >> > I think repeating is not good approach to fix race.
> >> >> > In your case, XFER_COMPLETE preceded data error and DTO didn't come?
> >> >> > It seems strange case.
> >> >> > I want to know actual error value if you can reproduce.
> >> >>
> >> >> XFER_COMPLETE didn't necessarily precede data error.  Imagine this scenario:
> >> >>
> >> >> 1. Check for data error: nope
> >> >> 2. Interrupt happens and we get a data error and immediately xfer complete
> >> >> 3. Check for xfer complete: yup
> >> >>
> >> >> That's the state that we are handling.
> >> >>
> >> >> The system that dw_mmc uses where the interrupt handler has no locking
> >> >> makes it incredibly difficult to get things right.  Can you propose an
> >> >> alternate fix that would avoid the race?
> >> > Thank you for detailed scenario.
> >> > You're right.
> >> > Have you consider using spin_lock() in interrupt handler?
> >> > Then, we'll need to change spin_lock() to spin_lock_irqsave() in tasklet func.
> >> > And other locks in driver may need to be adjusted properly.
> >>
> >> I have certainly considered it and I think it's the right way to go,
> >> but I believe that this would be a pretty massive change to the design
> >> of dw_mmc.  Someone appeared to try very hard not to use a spinlock in
> >> the interrupt handler and came up with the whole tasklet / pending
> >> events / completed events to deal with it.
> > Yes. It might be not small changes. Moreover, it needs to test heavily.
> > But it should be done before long. As we experienced, there are some race issue
> > in current way. Will you update this patch? Please let me know.
> 
> I would prefer not to, but I will if I need to.  As you say, adding a
> spinlock to the IRQ handler will need to be tested heavily and I'm not
> sure we're in the best position to do that.  I can send up the change
> to use a spinlock with some minimal testing if you want, though.
> ...of perhaps Yuvaraj can since he's the one who's been sending this
> patch upstream.
> 
> Note that the original change was quite hard to test.  The problem
> would show up on certain boards easier than others and would show up
> with certain revisions of the kernel but not with others.  In other
> words: it was heavily timing dependent.  The current patch as-is has
> been tested and shown to be stable for quite a while now.
> 
> 
> >> In this particular case it would be pretty easy to just add the
> >> spinlock around the data error / xfer complete checks, though once we
> >> have a spinlock here it seems like we'd want to start using it in
> >> other places.  It would be confusing if the interrupt handler grabbed
> >> a spinlock the whole time but then only used it to protect a single
> >> small check.
> >>
> >> I'm really curious, though, why this driver can't just use a threaded
> >> irq handler and eliminate the whole interrupt / tasklet split.  That
> >> seems saner in the long run.
> > I think dw_mmc conforms itself to typical driver's implementation,
> > which it may be about a choice of driver's design.
> 
> Yup, there are a lot of different valid ways to accomplish the same
> thing in the kernel.  ...but some of the ways drivers do things are
> historical and are not the suggested way of doing things anymore.  I'm
> still a kernel noob in many ways (despite a few years of working with
> it), so I may be mistaken.
> 
> To me it feels like the interrupt + tasklet is effectively reinventing
> threaded interrupts, but doing it in a custom way.

Although it may not be graceful way, this patch is plainly effective.
We can expect more in the next place.
I'll add minor comment in previous email.
Please check it.

Thanks,
Seungwon Jeon

--
To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux SoC Development]     [Linux Rockchip Development]     [Linux USB Development]     [Video for Linux]     [Linux Audio Users]     [Linux SCSI]     [Yosemite News]

  Powered by Linux