High order memory stuff within trace could introduce OOM, use kvmalloc instead.
Please find the bellowing for the call stack we run across in an android system. The scenario happens when traced_probes is woken up to get a large quantity of trace even if free memory is even higher than watermark_low.
traced_probes invoked oom-killer: gfp_mask=0x140c0c0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO), nodemask=(null), order=2, oom_score_adj=-1
traced_probes cpuset=system-background mems_allowed=0
CPU: 3 PID: 588 Comm: traced_probes Tainted: G W O 4.14.181 #1
Hardware name: Generic DT based system
(unwind_backtrace) from [<c010d824>] (show_stack+0x20/0x24)
(show_stack) from [<c0b2e174>] (dump_stack+0xa8/0xec)
(dump_stack) from [<c027d584>] (dump_header+0x9c/0x220)
(dump_header) from [<c027cfe4>] (oom_kill_process+0xc0/0x5c4)
(oom_kill_process) from [<c027cb94>] (out_of_memory+0x220/0x310)
(out_of_memory) from [<c02816bc>] (__alloc_pages_nodemask+0xff8/0x13a4)
(__alloc_pages_nodemask) from [<c02a6a1c>] (kmalloc_order+0x30/0x48)
(kmalloc_order) from [<c02a6a64>] (kmalloc_order_trace+0x30/0x118)
(kmalloc_order_trace) from [<c0223d7c>] (tracing_buffers_open+0x50/0xfc)
(tracing_buffers_open) from [<c02e6f58>] (do_dentry_open+0x278/0x34c)
(do_dentry_open) from [<c02e70d0>] (vfs_open+0x50/0x70)
(vfs_open) from [<c02f7c24>] (path_openat+0x5fc/0x169c)
(path_openat) from [<c02f75c4>] (do_filp_open+0x94/0xf8)
(do_filp_open) from [<c02e7650>] (do_sys_open+0x168/0x26c)
(do_sys_open) from [<c02e77bc>] (SyS_openat+0x34/0x38)
(SyS_openat) from [<c0108bc0>] (ret_fast_syscall+0x0/0x28)
Signed-off-by: Zhaoyang Huang <[email protected]>
---
kernel/trace/trace.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index ca1ee65..d4eb7ea 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -6891,7 +6891,7 @@ static int tracing_buffers_open(struct inode *inode, struct file *filp)
if (trace_array_get(tr) < 0)
return -ENODEV;
- info = kzalloc(sizeof(*info), GFP_KERNEL);
+ info = kvmalloc(sizeof(*info), GFP_KERNEL);
if (!info) {
trace_array_put(tr);
return -ENOMEM;
--
1.9.1
On Wed 29-07-20 19:48:29, Zhaoyang Huang wrote:
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index ca1ee65..d4eb7ea 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -6891,7 +6891,7 @@ static int tracing_buffers_open(struct inode *inode, struct file *filp)
> if (trace_array_get(tr) < 0)
> return -ENODEV;
>
> - info = kzalloc(sizeof(*info), GFP_KERNEL);
> + info = kvmalloc(sizeof(*info), GFP_KERNEL);
I am not familiar with the code but it is quite clear that the patch is
incomplete without matching kvfree call in the release path. You cannot
simply change k.malloc to kvmalloc without the pairing kvfree.
> if (!info) {
> trace_array_put(tr);
> return -ENOMEM;
> --
> 1.9.1
--
Michal Hocko
SUSE Labs