Message-ID: <20171030162314.yxi2fyli6omtphvw@lakrids.cambridge.arm.com>
Date:   Mon, 30 Oct 2017 16:23:15 +0000
From:   Mark Rutland <mark.rutland@....com>
To:     linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        will.deacon@....com, peterz@...radead.org, mingo@...hat.com
Subject: v4.14-rc{4,7} null pointer dereference in event_sched_out()

Hi,

As a heads-up, while fuzzing arm64 v4.14-rc{4,7} with Syzkaller, I hit a
KASAN splat in event_sched_out():

[  133.225742] ==================================================================
[  133.229374] BUG: KASAN: null-ptr-deref in event_sched_out.isra.47+0x428/0x580
[  133.230843] Read of size 4 at addr 0000000000000178 by task syz-executor0/6905
[  133.233151]
[  133.233664] CPU: 0 PID: 6905 Comm: syz-executor0 Not tainted 4.14.0-rc7-dirty #4
[  133.235750] Hardware name: linux,dummy-virt (DT)
[  133.236598] Call trace:
[  133.237081] [<ffff20000808fef8>] dump_backtrace+0x0/0x658
[  133.238073] [<ffff200008090570>] show_stack+0x20/0x30
[  133.239002] [<ffff2000091c22ec>] dump_stack+0xd0/0x124
[  133.239947] [<ffff200008349d1c>] kasan_report+0x104/0x310
[  133.240940] [<ffff2000083483f8>] __asan_load4+0x58/0xb0
[  133.242262] [<ffff200008271138>] event_sched_out.isra.47+0x428/0x580
[  133.243686] [<ffff2000082712c8>] __perf_remove_from_context+0x38/0xe0
[  133.244948] [<ffff200008265cf8>] event_function_call+0x1c8/0x258
[  133.246197] [<ffff20000826ad04>] perf_remove_from_context+0x54/0xf0
[  133.247514] [<ffff20000827f188>] SyS_perf_event_open+0x1528/0x18e0
[  133.248831] Exception stack(0xffff800038c5fec0 to 0xffff800038c60000)
[  133.250199] fec0: 0000000020b12f88 0000000000001af8 00000000ffffffff 0000000000000008
[  133.251843] fee0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
[  133.253503] ff00: 00000000000000f1 0000000000000000 0000000000405850 00000000003d0f00
[  133.255132] ff20: 0000ffff94514f60 00000000004ae890 0000000000000027 0000000000000001
[  133.256756] ff40: 0000000000000000 0000000000826000 0000000000000000 00000000004c0158
[  133.258392] ff60: 00000000ffffffff 0000000020b12f88 0000000000001af8 000000000046d290
[  133.260006] ff80: 00000000004aaba8 0000000000473af8 0000ffffe5360da0 0000000000000000
[  133.261629] ffa0: 0000ffff94514f60 0000ffff94514640 00000000004020fc 0000ffff94514640
[  133.263253] ffc0: 000000000042d034 00000000a0000000 0000000020b12f88 00000000000000f1
[  133.264886] ffe0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
[  133.266535] [<ffff200008084170>] el0_svc_naked+0x24/0x28
[  133.267648] ==================================================================

... which is triggered by the Syzkaller repro program at the end of this
email. I haven't yet come up with a C reproducer; sorry.

The PC seems to be the load of cpuctx->active_oncpu at the end of the
function, so it looks like cpuctx is NULL.

The system has (homogeneous) armv8_pmuv3, breakpoint, and software PMUs.

I initially hit this on v4.14-rc4, and can reproduce the issue on
v4.14-rc7. I haven't tried any other kernels yet.

I'll continue digging, unless someone else has already solved this.

Thanks,
Mark.

Syzkaller reproducer
---->8----
# {Threaded:true Collide:true Repeat:true Procs:1 Sandbox:none Fault:false FaultCall:-1 FaultNth:0 EnableTun:true UseTmpDir:true HandleSegv:true WaitRepeat:true Debug:false Repro:false}
mmap(&(0x7f0000000000/0xd3f000)=nil, 0xd3f000, 0x3, 0x32, 0xffffffffffffffff, 0x0)
r0 = perf_event_open(&(0x7f0000d15000-0x78)={0x1, 0x78, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x9, 0x34, 0x0, 0x8001, 0x0, 0x0, 0x0, 0x0, 0x0, 0x8, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x8000000, 0x0}, 0x0, 0xffffffff, 0xffffffffffffffff, 0x0)
mmap(&(0x7f0000d3f000/0x1000)=nil, 0x1000, 0x3, 0x32, 0xffffffffffffffff, 0x0)
r1 = perf_event_open(&(0x7f0000d15000-0x78)={0x1, 0x78, 0x0, 0x0, 0x0, 0x0, 0x0, 0x3bd4, 0x0, 0x0, 0x30, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffffffff80000001, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, 0x0, 0xffffffff, r0, 0x0)
add_key(&(0x7f00004fe000)="6465616400", &(0x7f0000d41000)={0x73, 0x79, 0x7a, 0x0, 0x0}, &(0x7f0000509000-0x19)="", 0x0, 0xfffffffffffffffd)
syz_emit_ethernet(0x32, &(0x7f0000b86000-0x36)={@...ote={[0xbb, 0xbb, 0xbb, 0xbb, 0xbb], 0x0}, @remote={[0xbb, 0xbb, 0xbb, 0xbb, 0xbb], 0x0}, [], {{0x200000000080a, @arp=@...eric={0x322, 0x8edf, 0x6, 0x0, 0xfffffffffffffffe, @empty=[0x0, 0x0, 0x0, 0x0, 0x0, 0x0], "", @random="dc1ce39913fb", "34f3b689bb48f1e976f2bf1b7cf2243b"}}}})
dup3(r1, r1, 0x80000)
r2 = gettid()
ioctl$sock_FIOSETOWN(0xffffffffffffffff, 0x8901, &(0x7f0000d40000)=0x0)
socket$inet6_udp(0xa, 0x2, 0x0)
perf_event_open(&(0x7f0000b13000-0x78)={0x0, 0x78, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x30, 0x0, 0x2, 0x0, 0x0, 0x94, 0x0, 0x0, 0x0, 0x0, 0x0, 0x10001, 0x0, 0x0, 0x0, 0x0, 0x0}, r2, 0xffffffff, r0, 0x0)

Powered by blists - more mailing lists