Oleg Nesterov found an interesting deadlock possibility:
sysrq_showregs_othercpus() does smp_call_function(showacpu) and showacpu() show_stack()->decode_address(). Now suppose that IPI interrupts the task holding read_lock(tasklist).
To fix this, blackfin should not grab the write_ variant of the tasklist lock, read_ one is enough.
Suggested-by: Oleg Nesterov oleg@redhat.com Signed-off-by: Anton Vorontsov anton.vorontsov@linaro.org --- arch/blackfin/kernel/trace.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/arch/blackfin/kernel/trace.c b/arch/blackfin/kernel/trace.c index d08f0e3..f7f7a18 100644 --- a/arch/blackfin/kernel/trace.c +++ b/arch/blackfin/kernel/trace.c @@ -29,7 +29,7 @@ void decode_address(char *buf, unsigned long address) { struct task_struct *p; struct mm_struct *mm; - unsigned long flags, offset; + unsigned long offset; struct rb_node *n;
#ifdef CONFIG_KALLSYMS @@ -113,7 +113,7 @@ void decode_address(char *buf, unsigned long address) * mappings of all our processes and see if we can't be a whee * bit more specific */ - write_lock_irqsave(&tasklist_lock, flags); + read_lock(&tasklist_lock); for_each_process(p) { struct task_struct *t;
@@ -186,7 +186,7 @@ __continue: sprintf(buf, "/* kernel dynamic memory */");
done: - write_unlock_irqrestore(&tasklist_lock, flags); + read_unlock(&tasklist_lock); }
#define EXPAND_LEN ((1 << CONFIG_DEBUG_BFIN_HWTRACE_EXPAND_LEN) * 256 - 1)