Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <CAHk-=wiqix9N5P0BXrSSOXjPZxMh=wDDRJ3sgf=hutoTUx0nZQ@mail.gmail.com>
 <20220905125637.GA2630968@roeck-us.net> <CAHk-=whh=0FG6r_YJ4_1pq07b=bqN8gTExU5T_ys-SVn0CRDtQ@mail.gmail.com>
 <752fba06-8ccc-d1e7-68d8-f141eece1230@roeck-us.net>
In-Reply-To: <752fba06-8ccc-d1e7-68d8-f141eece1230@roeck-us.net>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sat, 10 Sep 2022 17:32:30 -0400
Message-ID: <CAHk-=wj1obPoTu1AHj9Bd_BGYjdjDyPP+vT5WMj8eheb3A9WHw@mail.gmail.com>
Subject: Re: Linux 6.0-rc4
To:     Guenter Roeck <linux@roeck-us.net>,
        =?UTF-8?B?Q3PDs2vDoXMgQmVuY2U=?= <csokas.bence@prolan.hu>,
        "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Marc Kleine-Budde <mkl@pengutronix.de>,
        Francesco Dolcini <francesco.dolcini@toradex.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Let's bring in the guilty party and related people..

It looks like b353b241f1eb9 ("net: fec: Use a spinlock to guard
`fep->ptp_clk_on`") is not sufficient to fix the problems caused by
f79959220fa5 ("fec: Restart PPS after link state change"), and in fact
just causes more issues.

By now I suspect that both of those just have to be reverted.

Hmm?

              Linus

On Sat, Sep 10, 2022 at 2:13 PM Guenter Roeck <linux@roeck-us.net> wrote:
>
> On 9/10/22 10:44, Linus Torvalds wrote:
> >
> >> Revert "fec: Restart PPS after link state change"
> >>      revert f79959220fa5
> >>      Rationale:
> >>          The patch results in various tracebacks and crashes.
> >
> > Hmm. No revert, but does commit b353b241f1eb ("net: fec: Use a
> > spinlock to guard `fep->ptp_clk_on`") fix it?
> >
>
> No. The m68k crash is still there as of v6.0-rc4-284-gce888220d5c7. The above patch
> changes fec_enet_clk_enable(), but the crash is elsewhere - or at least now it is
> elsewhere.
>
> *** ILLEGAL INSTRUCTION ***   FORMAT=4
> Current process id is 1
> BAD KERNEL TRAP: 00000000
> PC: [<00000000>] 0x0
> SR: 2704  SP: (ptrval)  a2: 40829628
> d0: 00002700    d1: 00000000    d2: 40829420    d3: 00000000
> d4: 00000000    d5: 402e81aa    a0: 00000000    a1: 00000012
> Process swapper (pid: 1, task=(ptrval))
> Frame format=4 eff addr=400681c2 pc=00000000
> Stack from 40831cf0:
>          40829420 00000000 40832000 402e81aa 40bae000 00000008 40829420 40829000
>          401b0e44 40829628 40829420 00000000 40831e0c 00000200 401dce0a 40884a50
>          401b14d6 40829598 4082970e 40347ee0 401ad2d6 40829420 40347eea 00000000
>          40831e0c 402e81aa 40bae000 00000008 40347ee0 40829000 fffffff8 4082944e
>          40829000 408294bb 00000002 00000000 00000000 00000000 00000000 00000000
>          00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> Call Trace: [<401b0e44>] fec_ptp_gettime+0x28/0x68
>   [<401dce0a>] __alloc_skb+0xb0/0x24c
>   [<401b14d6>] fec_ptp_save_state+0x12/0x3e
>   [<401ad2d6>] fec_restart+0x5a/0x770
>   [<401ae4c0>] fec_probe+0x732/0xcee
> ...
>
> On top of that, I now also see
>
> [   22.043994] BUG: sleeping function called from invalid context at drivers/clk/imx/clk-pllv3.c:68
> [   22.044249] in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 272, name: ip
> [   22.044395] preempt_count: 1, expected: 0
> [   22.044485] 3 locks held by ip/272:
> [   22.044571]  #0: c1af1ce0 (rtnl_mutex){+.+.}-{3:3}, at: devinet_ioctl+0xbc/0x8f0
> [   22.044792]  #1: c40dc8e8 (&fep->tmreg_lock){....}-{2:2}, at: fec_enet_clk_enable+0x58/0x250
> [   22.045011]  #2: c1a71af8 (prepare_lock){+.+.}-{3:3}, at: clk_prepare_lock+0xc/0xd4
> [   22.045217] irq event stamp: 1460
> [   22.045295] hardirqs last  enabled at (1459): [<c1084cd8>] _raw_spin_unlock_irqrestore+0x50/0x64
> [   22.045454] hardirqs last disabled at (1460): [<c1084a98>] _raw_spin_lock_irqsave+0x64/0x68
> [   22.045623] softirqs last  enabled at (1444): [<c0e77268>] __dev_change_flags+0xa8/0x254
> [   22.045784] softirqs last disabled at (1442): [<c0e77238>] __dev_change_flags+0x78/0x254
> [   22.045944] CPU: 0 PID: 272 Comm: ip Tainted: G        W        N 6.0.0-rc4-00286-gb260ebd186c0 #1
> [   22.046113] Hardware name: Freescale i.MX7 Dual (Device Tree)
> [   22.046235]  unwind_backtrace from show_stack+0x10/0x14
> [   22.046360]  show_stack from dump_stack_lvl+0x68/0x90
> [   22.046477]  dump_stack_lvl from __might_resched+0x17c/0x284
> [   22.046604]  __might_resched from clk_pllv3_wait_lock+0x4c/0xcc
> [   22.046735]  clk_pllv3_wait_lock from clk_core_prepare+0xc4/0x328
> [   22.046866]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.046992]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047118]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047243]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047368]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047493]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047618]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047744]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047868]  clk_core_prepare from clk_core_prepare+0x50/0x328
> [   22.047992]  clk_core_prepare from clk_prepare+0x20/0x30
> [   22.048110]  clk_prepare from fec_enet_clk_enable+0x68/0x250
> [   22.048235]  fec_enet_clk_enable from fec_enet_open+0x3c/0x3b8
> [   22.048360]  fec_enet_open from __dev_open+0xec/0x1ac
> [   22.048476]  __dev_open from __dev_change_flags+0x1bc/0x254
> [   22.048603]  __dev_change_flags from dev_change_flags+0x14/0x44
> [   22.048730]  dev_change_flags from devinet_ioctl+0x88c/0x8f0
> [   22.048854]  devinet_ioctl from inet_ioctl+0x194/0x258
> [   22.048971]  inet_ioctl from sock_ioctl+0x4b0/0x5cc
> [   22.049083]  sock_ioctl from sys_ioctl+0x548/0xf18
> [   22.049198]  sys_ioctl from ret_fast_syscall+0x0/0x1c
>
> in various arm boot tests. This is new since -rc4.
>
> Guenter