Re: [PATCH] trace: fix race in perf_trace_buf initialization

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 17 Oct 2019 12:23:49 +0530
Prateek Sood <prsood@codeaurora.org> wrote:

> On 10/15/19 11:47 AM, Prateek Sood wrote:
> > [  943.034988] Unable to handle kernel paging request at virtual address 0000003106f2003c
> > [  943.043653] Mem abort info:
> > [  943.046679]   ESR = 0x96000045
> > [  943.050428]   Exception class = DABT (current EL), IL = 32 bits
> > [  943.056643]   SET = 0, FnV = 0
> > [  943.060168]   EA = 0, S1PTW = 0
> > [  943.063449] Data abort info:
> > [  943.066474]   ISV = 0, ISS = 0x00000045
> > [  943.070856]   CM = 0, WnR = 1
> > [  943.074016] user pgtable: 4k pages, 39-bit VAs, pgdp = ffffffc034b9b000
> > [  943.081446] [0000003106f2003c] pgd=0000000000000000, pud=0000000000000000
> > [  943.088862] Internal error: Oops: 96000045 [#1] PREEMPT SMP
> > [  943.141700] Process syz-executor (pid: 18393, stack limit = 0xffffffc093190000)
> > [  943.164146] pstate: 80400005 (Nzcv daif +PAN -UAO)
> > [  943.169119] pc : __memset+0x20/0x1ac
> > [  943.172831] lr : memset+0x3c/0x50
> > [  943.176269] sp : ffffffc09319fc50
> > 
> > [  943.557593]  __memset+0x20/0x1ac
> > [  943.560953]  perf_trace_buf_alloc+0x140/0x1a0
> > [  943.565472]  perf_trace_sys_enter+0x158/0x310
> > [  943.569985]  syscall_trace_enter+0x348/0x7c0
> > [  943.574413]  el0_svc_common+0x11c/0x368
> > [  943.578394]  el0_svc_handler+0x12c/0x198
> > [  943.582459]  el0_svc+0x8/0xc
> > 
> > In Ramdumps:
> > total_ref_count = 3
> > perf_trace_buf = (
> >     0x0 -> NULL,
> >     0x0 -> NULL,
> >     0x0 -> NULL,
> >     0x0 -> NULL)
> > 
> > event_call in perf_trace_sys_enter()
> > event_call = 0xFFFFFF900CB511D8 -> (
> >     list = (next = 0xFFFFFF900CB4E2E0, prev = 0xFFFFFF900CB512B0),
> >     class = 0xFFFFFF900CDC8308,
> >     name = 0xFFFFFF900CDDA1D8,
> >     tp = 0xFFFFFF900CDDA1D8,
> >     event = (
> >       node = (next = 0x0, pprev = 0xFFFFFF900CB80210),
> >       list = (next = 0xFFFFFF900CB512E0, prev = 0xFFFFFF900CB4E310),
> >       type = 21,
> >       funcs = 0xFFFFFF900CB51130),
> >     print_fmt = 0xFFFFFF900CB51150,
> >     filter = 0x0,
> >     mod = 0x0,
> >     data = 0x0,
> >     flags = 18,
> >     perf_refcount = 1,
> >     perf_events = 0xFFFFFF8DB8E54158,
> >     prog_array = 0x0,
> >     perf_perm = 0x0)
> > 
> > perf_events added on CPU0
> > (struct hlist_head *)(0xFFFFFF8DB8E54158+__per_cpu_offset[0]) -> (
> >     first = 0xFFFFFFC0980FD0E0 -> (
> >       next = 0x0,
> >       pprev = 0xFFFFFFBEBFD74158))
> > 
> > Could you please confirm:
> > 1) the race mentioned below exists or not.
> > 2) if following patch fixes it.
> > 
> >   
> >> 8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8>8  
> > 
> > A race condition exists while initialiazing perf_trace_buf from
> > perf_trace_init() and perf_kprobe_init().
> > 
> >       CPU0                                        CPU1
> > perf_trace_init()
> >   mutex_lock(&event_mutex)
> >     perf_trace_event_init()
> >       perf_trace_event_reg()
> >         total_ref_count == 0
> > 	buf = alloc_percpu()
> >         perf_trace_buf[i] = buf
> >         tp_event->class->reg() //fails       perf_kprobe_init()
> > 	goto fail                              perf_trace_event_init()
> >                                                  perf_trace_event_reg()
> >         fail:
> > 	  total_ref_count == 0
> > 
> >                                                    total_ref_count == 0
> >                                                    buf = alloc_percpu()
> >                                                    perf_trace_buf[i] = buf
> >                                                    tp_event->class->reg()
> >                                                    total_ref_count++
> > 
> >           free_percpu(perf_trace_buf[i])
> >           perf_trace_buf[i] = NULL
> > 
> > Any subsequent call to perf_trace_event_reg() will observe total_ref_count > 0,
> > causing the perf_trace_buf to be NULL always. This can result in perf_trace_buf
> > getting accessed from perf_trace_buf_alloc() without being initialized. Acquiring
> > event_mutex in perf_kprobe_init() before calling perf_trace_event_init() should
> > fix this race.
> > 
> > Signed-off-by: Prateek Sood <prsood@codeaurora.org>
> > ---
> >  kernel/trace/trace_event_perf.c | 4 ++++
> >  1 file changed, 4 insertions(+)
> > 
> > diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c
> > index 4629a61..48ee92c 100644
> > --- a/kernel/trace/trace_event_perf.c
> > +++ b/kernel/trace/trace_event_perf.c
> > @@ -272,9 +272,11 @@ int perf_kprobe_init(struct perf_event *p_event, bool is_retprobe)
> >  		goto out;
> >  	}
> >  
> > +	mutex_lock(&event_mutex);
> >  	ret = perf_trace_event_init(tp_event, p_event);
> >  	if (ret)
> >  		destroy_local_trace_kprobe(tp_event);
> > +	mutex_unlock(&event_mutex);
> >  out:
> >  	kfree(func);
> >  	return ret;
> > @@ -282,8 +284,10 @@ int perf_kprobe_init(struct perf_event *p_event, bool is_retprobe)
> >  
> >  void perf_kprobe_destroy(struct perf_event *p_event)
> >  {
> > +	mutex_lock(&event_mutex);
> >  	perf_trace_event_close(p_event);
> >  	perf_trace_event_unreg(p_event);
> > +	mutex_unlock(&event_mutex);
> >  
> >  	destroy_local_trace_kprobe(p_event->tp_event);
> >  }
> >   
> 
> Folks, 
> 
> Could you please help with this query/patch?

Added Peter Zijlstra who maintains the kernel side of perf.

-- Steve