Re: [PATCH 2/2] tracing: identify which executable object the userspace address belongs to

[Ingo Molnar <mingo@elte.hu>]

* Török Edwin <edwintorok@gmail.com> wrote:

> Impact: modify+improve the userstacktrace tracing visualization feature

> +static inline int seq_print_user_ip(struct trace_seq *s, struct mm_struct *mm,
> +				    unsigned long ip, unsigned long sym_flags)
> +{
> +	struct file *file = NULL;
> +	unsigned long vmstart = 0;
> +	int ret = 1;
> +
> +	if (mm) {
> +		const struct vm_area_struct *vma = find_vma(mm, ip);
> +		if (vma) {

Style: please put a newline after variable definition blocks.

> +			file = vma->vm_file;
> +			vmstart = vma->vm_start;

Bug: it's generally unsafe to look up a vma and use it without having 
done a down_read(&mm->mmap_sem). Another thread (of this ->mm) could 
go and modify it in parallel.

> +		}
> +	}
> +	if (file) {
> +		ret = trace_seq_path(s, &file->f_path);
> +		if (ret)
> +			ret = trace_seq_printf(s, "[+0x%lx]",
> +					ip - vmstart);

Style: that linebreak is unnecessary.

> +	if (trace_flags & TRACE_ITER_SYM_USEROBJ) {
> +		struct task_struct *task;
> +		/*
> +		 * we do the lookup on the thread group leader,
> +		 * since individual threads might have already quit!
> +		 */
> +		rcu_read_lock();
> +		task = find_task_by_vpid(entry->ent.tgid);
> +		rcu_read_unlock();
> +
> +		if (task)
> +			mm = get_task_mm(task);

Bug: it is unsafe to look up a task and then drop the RCU lock and use 
it - the task could go away the moment the RCU read-lock is dropped.

A safer sequence would be to get get_task_mm(task) reference inside 
the RCU critical section.

	Ingo