linux-kernel - Re: [PATCH] riscv: patch: remove lockdep assertion on lock text

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: <20220223071801.2gyv23lyf6gumpau@mail.google.com>
Date:   Wed, 23 Feb 2022 15:18:01 +0800
From:   Changbin Du <changbin.du@...il.com>
To:     Palmer Dabbelt <palmer@...belt.com>
Cc:     changbin.du@...il.com, Paul Walmsley <paul.walmsley@...ive.com>,
        aou@...s.berkeley.edu, linux-riscv@...ts.infradead.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] riscv: patch: remove lockdep assertion on lock text_mutex

On Wed, Apr 28, 2021 at 11:20:20PM -0700, Palmer Dabbelt wrote:
> On Fri, 16 Apr 2021 19:35:32 PDT (-0700), changbin.du@...il.com wrote:
> > The function patch_insn_write() expects that the text_mutex is already
> > held. There's a case that text_mutex is acquired by ftrace_run_update_code()
> > under syscall context but then patch_insn_write() will be executed under the
> > migration kthread context as we involves stop machine. So we should remove
> > the assertion, or it can cause warning storm in kernel message.
> > 
> > [  104.641978] ------------[ cut here ]------------
> > [  104.642327] WARNING: CPU: 0 PID: 13 at arch/riscv/kernel/patch.c:63 patch_insn_write+0x166/0x17c
> > [  104.643587] Modules linked in:
> > [  104.644691] CPU: 0 PID: 13 Comm: migration/0 Not tainted 5.12.0-rc7-00067-g9cdbf6467424 #102
> > [  104.644907] Hardware name: riscv-virtio,qemu (DT)
> > [  104.645068] Stopper: multi_cpu_stop+0x0/0x17e <- 0x0
> > [  104.645349] epc : patch_insn_write+0x166/0x17c
> > [  104.645467]  ra : patch_insn_write+0x162/0x17c
> > [  104.645534] epc : ffffffe0000059c6 ra : ffffffe0000059c2 sp : ffffffe002a33c70
> > [  104.645580]  gp : ffffffe0019e5518 tp : ffffffe002a232c0 t0 : ffffffe01295e8a8
> > [  104.645622]  t1 : 0000000000000001 t2 : 0000000000000000 s0 : ffffffe002a33cc0
> > [  104.645675]  s1 : ffffffe000007f72 a0 : 0000000000000000 a1 : ffffffffffffffff
> > [  104.645716]  a2 : 0000000000000001 a3 : 0000000000000000 a4 : 0000000000000001
> > [  104.645757]  a5 : ffffffe0799e45c8 a6 : 00000000000ca097 a7 : 0000000000000000
> > [  104.645798]  s2 : 0000000000000008 s3 : 0000000000000f72 s4 : ffffffe002a33ce0
> > [  104.645839]  s5 : 0000000000000f7a s6 : 0000000000000003 s7 : 0000000000000003
> > [  104.645880]  s8 : 0000000000000004 s9 : 0000000000000002 s10: 0000000000000000
> > [  104.645920]  s11: 0000000000000002 t3 : 0000000000000001 t4 : ffffffe000c615c8
> > [  104.645958]  t5 : 0000000000007fff t6 : 0000000000000380
> > [  104.645998] status: 0000000000000100 badaddr: 0000000000000000 cause: 0000000000000003
> > [  104.646081] Call Trace:
> > [  104.646147] [<ffffffe0000059c6>] patch_insn_write+0x166/0x17c
> > [  104.646280] [<ffffffe0000059ec>] patch_text_nosync+0x10/0x32
> > [  104.646317] [<ffffffe000007d0c>] ftrace_update_ftrace_func+0x74/0xac
> > [  104.646352] [<ffffffe0000d338a>] ftrace_modify_all_code+0x9c/0x144
> > [  104.646387] [<ffffffe0000d3444>] __ftrace_modify_code+0x12/0x1c
> > [  104.646420] [<ffffffe0000c4c2c>] multi_cpu_stop+0xa8/0x17e
> > [  104.646451] [<ffffffe0000c4474>] cpu_stopper_thread+0xb2/0x156
> > [  104.646489] [<ffffffe00003645e>] smpboot_thread_fn+0x102/0x1ea
> > [  104.646524] [<ffffffe0000311b0>] kthread+0x132/0x148
> > [  104.646556] [<ffffffe000002e02>] ret_from_exception+0x0/0x14
> > [  104.646657] ---[ end trace ccf71babb9de4d5b ]---
> > [  104.647444] ------------[ cut here ]------------
> > 
> > Signed-off-by: Changbin Du <changbin.du@...il.com>
> > ---
> >  arch/riscv/kernel/patch.c | 13 ++++++-------
> >  1 file changed, 6 insertions(+), 7 deletions(-)
> > 
> > diff --git a/arch/riscv/kernel/patch.c b/arch/riscv/kernel/patch.c
> > index 0b552873a577..6d2ed9c15065 100644
> > --- a/arch/riscv/kernel/patch.c
> > +++ b/arch/riscv/kernel/patch.c
> > @@ -49,19 +49,18 @@ static void patch_unmap(int fixmap)
> >  }
> >  NOKPROBE_SYMBOL(patch_unmap);
> > 
> > +
> > +/*
> > + * Before reaching here, it was expected to lock the text_mutex
> > + * already, so we don't need to give another lock here and could
> > + * ensure that it was safe between each cores.
> > + */
> >  static int patch_insn_write(void *addr, const void *insn, size_t len)
> >  {
> >  	void *waddr = addr;
> >  	bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
> >  	int ret;
> > 
> > -	/*
> > -	 * Before reaching here, it was expected to lock the text_mutex
> > -	 * already, so we don't need to give another lock here and could
> > -	 * ensure that it was safe between each cores.
> > -	 */
> > -	lockdep_assert_held(&text_mutex);
> > -
> >  	if (across_pages)
> >  		patch_map(addr + len, FIX_TEXT_POKE1);
> 
> Well, we definately need the lock -- otherwise concurrent callers will
> install conflicting maps for FIX_TEXT_POKE*, step on each other, and install
> the wrong patch.  I just sent out a patch to use an explicit lock here, as
> sharing text_mutex doesn't seem safe.

Just a note that this issue still exist on mainline. 
-- 
Cheers,
Changbin Du