linux-kernel - Re: reboot on wandboard fails with v4.14.67 (bisected to 2059e527a6)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <d8dd8d02-b8ed-91fd-bcc0-87dedc5998ae@prevas.dk>
Date:   Wed, 29 Aug 2018 15:53:57 +0200
From:   Rasmus Villemoes <rasmus.villemoes@...vas.dk>
To:     stable@...r.kernel.org, Fabio Estevam <fabio.estevam@....com>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        Sean Nyekjær <Sean.Nyekjaer@...vas.dk>,
        Andrey Smirnov <andrew.smirnov@...il.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Shawn Guo <shawnguo@...nel.org>
Subject: Re: reboot on wandboard fails with v4.14.67 (bisected to 2059e527a6)

On 2018-08-29 12:55, Rasmus Villemoes wrote:
> On 2018-08-29 11:55, Rasmus Villemoes wrote:
>> We're using imx_v6_v7_defconfig on our Wandboards. After upgrading to
>> v4.14.67, reboot no longer works (or, well, takes a very long time when
>> the watchdog is configured).
>>
>> v4.14.66 works fine, the breakage bisects to
>> 2059e527a659cf16d6bb709f1c8509f7a7623fc4 (ARM: imx_v6_v7_defconfig:
>> Select ULPI support), and reverting that on top of v4.14.67 again works.
> 
> FWIW, v4.18 (the mainline release containing the upstream commit
> 157bcc0609) works fine, while v4.17.19 which also has that commit
> backported (as df06ca1f56) fails. So I'm guessing there must be some
> other commits in v4.18 that happen to make imx_v6_v7_defconfig work.

OK, so v4.18-rc4 is also broken. Bisecting between that and v4.18 points
to a930d8bd94d (usb: chipidea: Always build ULPI code) introduced in
-rc7. The Makefile/Kconfig parts of that commit should not make any
difference wrt. 2059e527a6, since that commit explicitly set
USB_CHIPIDEA_ULPI, so we compile the same TUs with or without
a930d8bd94d. However, a930d8bd94d also adds this hunk

diff --git a/drivers/usb/chipidea/ulpi.c b/drivers/usb/chipidea/ulpi.c
index 6da42dcd2888..dfec07e8ae1d 100644
--- a/drivers/usb/chipidea/ulpi.c
+++ b/drivers/usb/chipidea/ulpi.c
@@ -95,6 +95,9 @@ int ci_ulpi_resume(struct ci_hdrc *ci)
 {
        int cnt = 100000;

+       if (ci->platdata->phy_mode != USBPHY_INTERFACE_MODE_ULPI)
+               return 0;
+
        while (cnt-- > 0) {
                if (hw_read(ci, OP_ULPI_VIEWPORT, ULPI_SYNC_STATE))
                        return 0;

and indeed just adding that on top of v4.14.67 makes it work again.

I don't know what the best fix for -stable is; the hunk above is the
minimal fix, but does not correspond to a mainline commit. a930d8bd94d
does cherrypick cleanly to v4.14.67, so might be the better option.

Rasmus