[U-Boot] [PATCH 1/2] WIP: Disable SCSI on x86

[Bin Meng <bmeng.cn@gmail.com>]

Hi Stefan,

On Tue, Jun 27, 2017 at 2:42 AM, Simon Glass <sjg@chromium.org> wrote:
> Hi,
>
> On 26 June 2017 at 03:11, Bin Meng <bmeng.cn@gmail.com> wrote:
>>
>> Hi Stefan,
>>
>> On Mon, Jun 26, 2017 at 3:29 PM, Stefan Roese <sr@denx.de> wrote:
>> > Hi Bin,
>> >
>> >
>> > On 24.06.2017 14:08, Bin Meng wrote:
>> >>
>> >> Hi Stefan,
>> >>
>> >> On Sat, Jun 24, 2017 at 6:29 PM, Stefan Roese <sr@denx.de> wrote:
>> >>>
>> >>> Hi Simon, Hi Bin,
>> >>>
>> >>>
>> >>> On 12.06.2017 05:53, Simon Glass wrote:
>> >>>>
>> >>>>
>> >>>> Hi Bin,
>> >>>>
>> >>>> On 9 June 2017 at 21:47, Bin Meng <bmeng.cn@gmail.com> wrote:
>> >>>>>
>> >>>>>
>> >>>>> Hi Simon,
>> >>>>>
>> >>>>> On Fri, Jun 9, 2017 at 8:27 PM, Simon Glass <sjg@chromium.org> wrote:
>> >>>>>>
>> >>>>>>
>> >>>>>> Hi Bin,
>> >>>>>>
>> >>>>>> On 1 June 2017 at 17:04, Bin Meng <bmeng.cn@gmail.com> wrote:
>> >>>>>>>
>> >>>>>>>
>> >>>>>>> Hi Simon,
>> >>>>>>>
>> >>>>>>> On Thu, Jun 1, 2017 at 9:14 PM, Simon Glass <sjg@chromium.org> wrote:
>> >>>>>>>>
>> >>>>>>>>
>> >>>>>>>> Hi Bin,
>> >>>>>>>>
>> >>>>>>>> On 1 June 2017 at 01:27, Bin Meng <bmeng.cn@gmail.com> wrote:
>> >>>>>>>>>
>> >>>>>>>>>
>> >>>>>>>>>
>> >>>>>>>>> Hi Simon,
>> >>>>>>>>>
>> >>>>>>>>> On Wed, May 17, 2017 at 8:42 AM, Bin Meng <bmeng.cn@gmail.com>
>> >>>>>>>>> wrote:
>> >>>>>>>>>>
>> >>>>>>>>>>
>> >>>>>>>>>> Hi Simon,
>> >>>>>>>>>>
>> >>>>>>>>>> On Sat, May 13, 2017 at 9:11 AM, Simon Glass <sjg@chromium.org>
>> >>>>>>>>>> wrote:
>> >>>>>>>>>>>
>> >>>>>>>>>>>
>> >>>>>>>>>>> Hi Bin,
>> >>>>>>>>>>>
>> >>>>>>>>>>> On 8 May 2017 at 01:05, Bin Meng <bmeng.cn@gmail.com> wrote:
>> >>>>>>>>>>>>
>> >>>>>>>>>>>>
>> >>>>>>>>>>>> Hi Simon,
>> >>>>>>>>>>>>
>> >>>>>>>>>>>> On Mon, Apr 10, 2017 at 8:38 AM, Simon Glass <sjg@chromium.org>
>> >>>>>>>>>>>> wrote:
>> >>>>>>>>>>>>>
>> >>>>>>>>>>>>>
>> >>>>>>>>>>>>> This is not to be applied. It is needed to test using
>> >>>>>>>>>>>>> CONFIG_DM_MMC on x86
>> >>>>>>>>>>>>> since SCSI is broken with this setup.
>> >>>>>>>>>>>>>
>> >>>>>>>>>>>>> Signed-off-by: Simon Glass <sjg@chromium.org>
>> >>>>>>>>>>>>> ---
>> >>>>>>>>>>>>
>> >>>>>>>>>>>>
>> >>>>>>>>>>>>
>> >>>>>>>>>>>> When you said "SCSI is broken with this setup", did you mean GP#
>> >>>>>>>>>>>> exception when boot up with a hard disk like below?
>> >>>>>>>>>>>
>> >>>>>>>>>>>
>> >>>>>>>>>>>
>> >>>>>>>>>>> Yes that's right.
>> >>>>>>>>>>>
>> >>>>>>>>>>
>> >>>>>>>>>> Do you plan to work on a patch soon?
>> >>>>>>>>>>
>> >>>>>>>>>
>> >>>>>>>>> I tracked down the exception happens at
>> >>>>>>>>> part_test_efi()->block_dread() function:
>> >>>>>>>>>
>> >>>>>>>>> if (!ops->read)
>> >>>>>>>>>       return -ENOSYS;
>> >>>>>>>>>
>> >>>>>>>>> Where ops points to nowhere, because underlying SATA device does
>> >>>>>>>>> not
>> >>>>>>>>> have a DM driver associated.
>> >>>>>>>>>
>> >>>>>>>>> Could you please fix this?
>> >>>>>>>>
>> >>>>>>>>
>> >>>>>>>>
>> >>>>>>>> I've been hoping that someone else will take on this side of things,
>> >>>>>>>> but perhaps I am optimistic. I'll see if I can take a look.
>> >>>>>>>
>> >>>>>>>
>> >>>>>>>
>> >>>>>>> Thanks. I wanted to fix this myself however I have been busy on some
>> >>>>>>> other stuff.
>> >>>>>>
>> >>>>>>
>> >>>>>>
>> >>>>>> I've sent a series which I think puts SCSI and SATA in a better
>> >>>>>> position with respect to driver model. However even with that there is
>> >>>>>> more work to do.
>> >>>>>>
>> >>>>>
>> >>>>> Thanks for working on this! I will take a look.
>> >>>>>
>> >>>>> If we don't get enough time to sort out everything, we will have to
>> >>>>> revert previous MMC DM conversion patch and leave that to next
>> >>>>> release.
>> >>>>
>> >>>>
>> >>>>
>> >>>> Yes, let's see how it goes. I probably have time for one more spin if
>> >>>> it is soon...but the last patch needs work.
>> >>>
>> >>>
>> >>>
>> >>> As Andreas has pointed out here:
>> >>>
>> >>> https://lists.denx.de/pipermail/u-boot/2017-June/296337.html
>> >>>
>> >>> SCSI seems to be broken also on Armada 37xx (and perhaps other
>> >>> platforms as well). Is it correct to assume, that this x86 SCSI
>> >>> thread is about the same issue?
>> >>>
>> >>> I know that you are working on DM SCSI support, which is great of
>> >>> course. But do you have an idea, if and how this could be fixed for
>> >>> this upcoming release?
>> >>
>> >>
>> >> Simon already sent a series to convert SCSI to DM and applied to
>> >> u-boot-dm. I also sent a series to fix some additional issues [1].
>> >>
>> >> Can you or Andreas try on top of u-boot-dm plus my patches to see if
>> >> the issue is fixed?
>> >>
>> >> [1] http://patchwork.ozlabs.org/patch/777312/
>> >
>> >
>> > I tested with this version and "scsi reset" still reboots the
>> > board. Please note that SCSI / AHCI probably needs some additional
>> > work, as its currently supported via this driver in arch/arm
>> > which needs to moved to the correct driver directory:
>> >
>> > arch/arm/mach-mvebu/sata.c
>> >
>> > It probably also needs some work to function correctly with these
>> > new DM SCSI patches. But all this seems a bit late in this release
>> > cycle. Do we have a "solution" to fix this issue for this upcoming
>> > release?
>> >
>>
>> Did the failure happen at the place part_test_efi()->block_dread() function?
>>
>> if (!ops->read)
>>     return -ENOSYS;
>>
>> Where ops points to nowhere ..
>>
>> If so, I think the SATA driver needs to be converted to DM.
>>
>> But I don't know if a simple solution exists for now on your board. I
>> noticed that Andreas said v2017.01 works but v2017.03 is failing. So
>> it has been broken for some time. But Simon's series aims to resolve
>> the issues seen on x86, which is an intermediate issue in this release
>> cycle.
>
> I sent a pull request (forgetting to cc the list) for this but it was
> too close to RC2 so Tom rejected it. It is sitting in dm/master if you
> want to look.
>
> At this stage I don't have a good answer. Even my series doesn't
> complete the SATA conversion to DM. We still need AHCI to have proper
> operations and convert the rest of the drivers. I was hoping that
> might be easier once the powerpc removals are complete.

Since the pull request of u-boot-dm was rejected for this release, I
sent a revert patch for x86 [1] to fix x86 SCSI issues.

Is Armada 37xx converted to use DM MMC? If so, maybe a quick fix is to
revert the Armada 37xx board back to use non-DM MMC, like x86 does?

[1] http://patchwork.ozlabs.org/patch/780913/

Regards,
Bin