https://github.com/torvalds/linux
Revision 6376360ecbe525a9c17b3d081dfd88ba3e4ed65b authored by Naoya Horiguchi on 01 February 2019, 22:21:08 UTC, committed by Linus Torvalds on 01 February 2019, 23:46:23 UTC
Currently memory_failure() is racy against process's exiting, which results in kernel crash by null pointer dereference. The root cause is that memory_failure() uses force_sig() to forcibly kill asynchronous (meaning not in the current context) processes. As discussed in thread https://lkml.org/lkml/2010/6/8/236 years ago for OOM fixes, this is not a right thing to do. OOM solves this issue by using do_send_sig_info() as done in commit d2d393099de2 ("signal: oom_kill_task: use SEND_SIG_FORCED instead of force_sig()"), so this patch is suggesting to do the same for hwpoison. do_send_sig_info() properly accesses to siglock with lock_task_sighand(), so is free from the reported race. I confirmed that the reported bug reproduces with inserting some delay in kill_procs(), and it never reproduces with this patch. Note that memory_failure() can send another type of signal using force_sig_mceerr(), and the reported race shouldn't happen on it because force_sig_mceerr() is called only for synchronous processes (i.e. BUS_MCEERR_AR happens only when some process accesses to the corrupted memory.) Link: http://lkml.kernel.org/r/20190116093046.GA29835@hori1.linux.bs1.fc.nec.co.jp Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Reported-by: Jane Chu <jane.chu@oracle.com> Reviewed-by: Dan Williams <dan.j.williams@intel.com> Reviewed-by: William Kucharski <william.kucharski@oracle.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent 0d0c8de
Tip revision: 6376360ecbe525a9c17b3d081dfd88ba3e4ed65b authored by Naoya Horiguchi on 01 February 2019, 22:21:08 UTC
mm: hwpoison: use do_send_sig_info() instead of force_sig()
mm: hwpoison: use do_send_sig_info() instead of force_sig()
Tip revision: 6376360
blk-stat.h
/* SPDX-License-Identifier: GPL-2.0 */
#ifndef BLK_STAT_H
#define BLK_STAT_H
#include <linux/kernel.h>
#include <linux/blkdev.h>
#include <linux/ktime.h>
#include <linux/rcupdate.h>
#include <linux/timer.h>
/**
* struct blk_stat_callback - Block statistics callback.
*
* A &struct blk_stat_callback is associated with a &struct request_queue. While
* @timer is active, that queue's request completion latencies are sorted into
* buckets by @bucket_fn and added to a per-cpu buffer, @cpu_stat. When the
* timer fires, @cpu_stat is flushed to @stat and @timer_fn is invoked.
*/
struct blk_stat_callback {
/*
* @list: RCU list of callbacks for a &struct request_queue.
*/
struct list_head list;
/**
* @timer: Timer for the next callback invocation.
*/
struct timer_list timer;
/**
* @cpu_stat: Per-cpu statistics buckets.
*/
struct blk_rq_stat __percpu *cpu_stat;
/**
* @bucket_fn: Given a request, returns which statistics bucket it
* should be accounted under. Return -1 for no bucket for this
* request.
*/
int (*bucket_fn)(const struct request *);
/**
* @buckets: Number of statistics buckets.
*/
unsigned int buckets;
/**
* @stat: Array of statistics buckets.
*/
struct blk_rq_stat *stat;
/**
* @fn: Callback function.
*/
void (*timer_fn)(struct blk_stat_callback *);
/**
* @data: Private pointer for the user.
*/
void *data;
struct rcu_head rcu;
};
struct blk_queue_stats *blk_alloc_queue_stats(void);
void blk_free_queue_stats(struct blk_queue_stats *);
void blk_stat_add(struct request *rq, u64 now);
/* record time/size info in request but not add a callback */
void blk_stat_enable_accounting(struct request_queue *q);
/**
* blk_stat_alloc_callback() - Allocate a block statistics callback.
* @timer_fn: Timer callback function.
* @bucket_fn: Bucket callback function.
* @buckets: Number of statistics buckets.
* @data: Value for the @data field of the &struct blk_stat_callback.
*
* See &struct blk_stat_callback for details on the callback functions.
*
* Return: &struct blk_stat_callback on success or NULL on ENOMEM.
*/
struct blk_stat_callback *
blk_stat_alloc_callback(void (*timer_fn)(struct blk_stat_callback *),
int (*bucket_fn)(const struct request *),
unsigned int buckets, void *data);
/**
* blk_stat_add_callback() - Add a block statistics callback to be run on a
* request queue.
* @q: The request queue.
* @cb: The callback.
*
* Note that a single &struct blk_stat_callback can only be added to a single
* &struct request_queue.
*/
void blk_stat_add_callback(struct request_queue *q,
struct blk_stat_callback *cb);
/**
* blk_stat_remove_callback() - Remove a block statistics callback from a
* request queue.
* @q: The request queue.
* @cb: The callback.
*
* When this returns, the callback is not running on any CPUs and will not be
* called again unless readded.
*/
void blk_stat_remove_callback(struct request_queue *q,
struct blk_stat_callback *cb);
/**
* blk_stat_free_callback() - Free a block statistics callback.
* @cb: The callback.
*
* @cb may be NULL, in which case this does nothing. If it is not NULL, @cb must
* not be associated with a request queue. I.e., if it was previously added with
* blk_stat_add_callback(), it must also have been removed since then with
* blk_stat_remove_callback().
*/
void blk_stat_free_callback(struct blk_stat_callback *cb);
/**
* blk_stat_is_active() - Check if a block statistics callback is currently
* gathering statistics.
* @cb: The callback.
*/
static inline bool blk_stat_is_active(struct blk_stat_callback *cb)
{
return timer_pending(&cb->timer);
}
/**
* blk_stat_activate_nsecs() - Gather block statistics during a time window in
* nanoseconds.
* @cb: The callback.
* @nsecs: Number of nanoseconds to gather statistics for.
*
* The timer callback will be called when the window expires.
*/
static inline void blk_stat_activate_nsecs(struct blk_stat_callback *cb,
u64 nsecs)
{
mod_timer(&cb->timer, jiffies + nsecs_to_jiffies(nsecs));
}
static inline void blk_stat_deactivate(struct blk_stat_callback *cb)
{
del_timer_sync(&cb->timer);
}
/**
* blk_stat_activate_msecs() - Gather block statistics during a time window in
* milliseconds.
* @cb: The callback.
* @msecs: Number of milliseconds to gather statistics for.
*
* The timer callback will be called when the window expires.
*/
static inline void blk_stat_activate_msecs(struct blk_stat_callback *cb,
unsigned int msecs)
{
mod_timer(&cb->timer, jiffies + msecs_to_jiffies(msecs));
}
void blk_rq_stat_add(struct blk_rq_stat *, u64);
void blk_rq_stat_sum(struct blk_rq_stat *, struct blk_rq_stat *);
void blk_rq_stat_init(struct blk_rq_stat *);
#endif
Computing file changes ...