Re: [PATCH] sunrpc: Fix trace events to store data in the struct

From: James Hogan
Date: Tue Feb 24 2015 - 08:46:10 EST


Hi Trond,

On 24/02/15 13:36, Trond Myklebust wrote:
> On Tue, Feb 24, 2015 at 6:47 AM, James Hogan <james.hogan@xxxxxxxxxx> wrote:
>> Commit 83a712e0afef ("sunrpc: add some tracepoints around enqueue and
>> dequeue of svc_xprt") merged in v3.19-rc1 added some new trace events,
>> however a couple of them printed data from dereferenced pointers rather
>> than storing the data in the struct. In general this isn't safe as the
>> print may not happen until later when the data may have changed or been
>> freed, and nor is it portable as userland won't have access to that
>> other data in order to interpret the trace data itself.
>>
>> Fix by copying the data into the struct and printing from there.
>>
>> Fixes: 83a712e0afef ("sunrpc: add some tracepoints around enqueue ...")
>> Signed-off-by: James Hogan <james.hogan@xxxxxxxxxx>
>> Cc: Jeff Layton <jlayton@xxxxxxxxxxxxxxx>
>> Cc: J. Bruce Fields <bfields@xxxxxxxxxx>
>> Cc: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
>> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
>> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
>> Cc: <stable@xxxxxxxxxxxxxxx> # v3.19+
>> ---
>> Build tested only. Perhaps somebody familiar with the code could give it
>> a spin to sanity check the trace output.
>> ---
>> include/trace/events/sunrpc.h | 22 +++++++++++++++-------
>> 1 file changed, 15 insertions(+), 7 deletions(-)
>>
>> diff --git a/include/trace/events/sunrpc.h b/include/trace/events/sunrpc.h
>> index b9c1dc6c825a..47dfcaebfaaf 100644
>> --- a/include/trace/events/sunrpc.h
>> +++ b/include/trace/events/sunrpc.h
>> @@ -503,18 +503,22 @@ TRACE_EVENT(svc_xprt_do_enqueue,
>>
>> TP_STRUCT__entry(
>> __field(struct svc_xprt *, xprt)
>> - __field(struct svc_rqst *, rqst)
>> + __field_struct(struct sockaddr_storage, ss)
>> + __field(unsigned long, flags);
>> + __field(int, pid)
>> ),
>>
>> TP_fast_assign(
>> __entry->xprt = xprt;
>> - __entry->rqst = rqst;
>> + xprt ? memcpy(&__entry->ss, &xprt->xpt_remote, sizeof(__entry->ss)) : memset(&__entry->ss, 0, sizeof(__entry->ss));
>
> How could xprt ever be NULL here, and even if it was, why the esoteric
> C instead of a simple 'if' statement?

Yeh, I had a straight forward unconditional assignment before, but I
changed it purely for consistency with the svc_xprt_dequeue trace event.

I don't pretend to understand the details of what is being traced
though, so I'm happy to change it if required.

Thanks
James

>
>> + __entry->flags = xprt ? xprt->xpt_flags : 0;
>> + __entry->pid = rqst ? rqst->rq_task->pid : 0;
>> ),
>>
>> TP_printk("xprt=0x%p addr=%pIScp pid=%d flags=%s", __entry->xprt,
>> - (struct sockaddr *)&__entry->xprt->xpt_remote,
>> - __entry->rqst ? __entry->rqst->rq_task->pid : 0,
>> - show_svc_xprt_flags(__entry->xprt->xpt_flags))
>> + (struct sockaddr *)&__entry->ss,
>> + __entry->pid,
>> + show_svc_xprt_flags(__entry->flags))
>> );
>>
>> TRACE_EVENT(svc_xprt_dequeue,
>> @@ -562,17 +566,21 @@ TRACE_EVENT(svc_handle_xprt,
>>
>> TP_STRUCT__entry(
>> __field(struct svc_xprt *, xprt)
>> + __field_struct(struct sockaddr_storage, ss)
>> + __field(unsigned long, flags);
>> __field(int, len)
>> ),
>>
>> TP_fast_assign(
>> __entry->xprt = xprt;
>> + xprt ? memcpy(&__entry->ss, &xprt->xpt_remote, sizeof(__entry->ss)) : memset(&__entry->ss, 0, sizeof(__entry->ss));
>
> Ditto.
>
>> + __entry->flags = xprt ? xprt->xpt_flags : 0;
>> __entry->len = len;
>> ),
>>
>> TP_printk("xprt=0x%p addr=%pIScp len=%d flags=%s", __entry->xprt,
>> - (struct sockaddr *)&__entry->xprt->xpt_remote, __entry->len,
>> - show_svc_xprt_flags(__entry->xprt->xpt_flags))
>> + (struct sockaddr *)&__entry->ss, __entry->len,
>> + show_svc_xprt_flags(__entry->flags))
>> );
>> #endif /* _TRACE_SUNRPC_H */
>>
>> --
>> 2.0.5
>>
>
>
>

Attachment: signature.asc
Description: OpenPGP digital signature