Re: Regression on Linux 6.0-rc1

From: Anup Patel <anup@brainfault.org>
To: Ron Economos <re@w6rz.net>
Cc: linux-riscv@lists.infradead.org, conor.dooley@microchip.com,
	 palmer@rivosinc.com, daolu@rivosinc.com, heiko@sntech.de
Subject: Re: Regression on Linux 6.0-rc1
Date: Tue, 16 Aug 2022 21:14:41 +0530	[thread overview]
Message-ID: <CAAhSdy2T8txKiDYYfwR3NpK+0jR1oHnaU7ZX5h+g+0y9obHxjQ@mail.gmail.com> (raw)
In-Reply-To: <728ecbd5-975a-168e-efab-3c0030be21d5@w6rz.net>

On Tue, Aug 16, 2022 at 6:29 PM Ron Economos <re@w6rz.net> wrote:
>
> The following Oops occurs on Linux 6.0-rc1 on the HiFive Unmatched. This
> is with gcc 11.2.0 and binutils 2.38 (from
> https://github.com/riscv-collab/riscv-gnu-toolchain).
>
> Begin: Loading essential drivers ... [   14.774748] Unable to handle
> kernel paging request at virtual address ffffffff813cdc90
> [   14.781976] Oops [#1]
> [   14.784163] Modules linked in: raid6_pq(+) libcrc32c raid1 raid0
> multipath linear da9063_regulator
> [   14.800763] CPU: 1 PID: 248 Comm: modprobe Not tainted 6.0.0-rc1 #1
> [   14.806747] Hardware name: SiFive HiFive Unmatched A00 (DT)
> [   14.812307] epc : jump_label_module_notify+0x1bc/0x270
> [   14.817430]  ra : jump_label_module_notify+0x19a/0x270
> [   14.822555] epc : ffffffff801b804c ra : ffffffff801b802a sp :
> ffffffc80424bc0
> [   14.829766]  gp : ffffffff81a2d7b8 tp : ffffffd895ae1980 t0 :
> ffffffd88067e20
> [   14.836975]  t1 : 6f1d98585898183f t2 : ffffffdbfed19e58 s0 :
> ffffffc80424bc0
> [   14.844185]  s1 : ffffffff813cdc88 a0 : 0000000000000000 a1 :
> 000000000000000
> [   14.851394]  a2 : 0000000000000000 a3 : ffffffff01c77000 a4 :
> 000000000000000
> [   14.858602]  a5 : 0000000000000000 a6 : 0000000000ff0000 a7 :
> 000000000000021
> [   14.865812]  s2 : ffffffff01c73000 s3 : ffffffff01c740c0 s4 :
> ffffffd88067e80
> [   14.873021]  s5 : ffffffff01c73010 s6 : 0000000000000dc0 s7 :
> ffffffff813cdf0
> [   14.880230]  s8 : ffffffd88067e282 s9 : ffffffff01c73008 s10:
> 000000000000002
> [   14.887439]  s11: ffffffc80424bd50 t3 : e26cb9592a0e8a69 t4 :
> 000000800000000
> [   14.894649]  t5 : 000000ff00000000 t6 : ffffffd8819ac4a0
> [   14.899945] status: 0000000200000120 badaddr: ffffffff813cdc90 cause:
> 000000f
> [   14.907853] [<ffffffff8004055a>]
> blocking_notifier_call_chain_robust+0x72/0x8
> [   14.915061] [<ffffffff800a1dfc>] load_module+0x17bc/0x1e98
> [   14.920531] [<ffffffff800a2718>] __do_sys_finit_module+0x98/0x100
> [   14.926611] [<ffffffff800a27a4>] sys_finit_module+0x24/0x30
> [   14.932170] [<ffffffff80003ca6>] ret_from_syscall+0x0/0x2
> [   14.937680] ---[ end trace 0000000000000000 ]---
> Segmentation fault
>
> Manually bisected to this commit:
>
> arch/riscv: add Zihintpause support
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8eb060e10185cfc97ef0200d197ec246ba0f9f8c

We see this crash for any loadable module which uses
cpu_relax(). In fact, KVM RISC-V module loading is also
broken.

Regards,
Anup

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv