diff options
author | Peter Zijlstra <a.p.zijlstra@chello.nl> | 2009-09-23 18:57:03 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2009-09-24 10:21:01 -0400 |
commit | ba0a6c9f6fceed11c6a99e8326f0477fe383e6b5 (patch) | |
tree | 5c9967735f19b7b3332de90cae2cdd6c4457d30c /fs/fcntl.c | |
parent | 06f1631a1671bce2d59d58cb1debdf23e1803ebf (diff) |
fcntl: add F_[SG]ETOWN_EX
In order to direct the SIGIO signal to a particular thread of a
multi-threaded application we cannot, like suggested by the manpage, put a
TID into the regular fcntl(F_SETOWN) call. It will still be send to the
whole process of which that thread is part.
Since people do want to properly direct SIGIO we introduce F_SETOWN_EX.
The need to direct SIGIO comes from self-monitoring profiling such as with
perf-counters. Perf-counters uses SIGIO to notify that new sample data is
available. If the signal is delivered to the same task that generated the
new sample it can augment that data by inspecting the task's user-space
state right after it returns from the kernel. This is esp. convenient
for interpreted or virtual machine driven environments.
Both F_SETOWN_EX and F_GETOWN_EX take a pointer to a struct f_owner_ex
as argument:
struct f_owner_ex {
int type;
pid_t pid;
};
Where type is one of F_OWNER_TID, F_OWNER_PID or F_OWNER_GID.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Tested-by: stephane eranian <eranian@googlemail.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'fs/fcntl.c')
-rw-r--r-- | fs/fcntl.c | 108 |
1 files changed, 100 insertions, 8 deletions
diff --git a/fs/fcntl.c b/fs/fcntl.c index 50a988993262..fc089f2f7f56 100644 --- a/fs/fcntl.c +++ b/fs/fcntl.c | |||
@@ -263,6 +263,79 @@ pid_t f_getown(struct file *filp) | |||
263 | return pid; | 263 | return pid; |
264 | } | 264 | } |
265 | 265 | ||
266 | static int f_setown_ex(struct file *filp, unsigned long arg) | ||
267 | { | ||
268 | struct f_owner_ex * __user owner_p = (void * __user)arg; | ||
269 | struct f_owner_ex owner; | ||
270 | struct pid *pid; | ||
271 | int type; | ||
272 | int ret; | ||
273 | |||
274 | ret = copy_from_user(&owner, owner_p, sizeof(owner)); | ||
275 | if (ret) | ||
276 | return ret; | ||
277 | |||
278 | switch (owner.type) { | ||
279 | case F_OWNER_TID: | ||
280 | type = PIDTYPE_MAX; | ||
281 | break; | ||
282 | |||
283 | case F_OWNER_PID: | ||
284 | type = PIDTYPE_PID; | ||
285 | break; | ||
286 | |||
287 | case F_OWNER_GID: | ||
288 | type = PIDTYPE_PGID; | ||
289 | break; | ||
290 | |||
291 | default: | ||
292 | return -EINVAL; | ||
293 | } | ||
294 | |||
295 | rcu_read_lock(); | ||
296 | pid = find_vpid(owner.pid); | ||
297 | if (owner.pid && !pid) | ||
298 | ret = -ESRCH; | ||
299 | else | ||
300 | ret = __f_setown(filp, pid, type, 1); | ||
301 | rcu_read_unlock(); | ||
302 | |||
303 | return ret; | ||
304 | } | ||
305 | |||
306 | static int f_getown_ex(struct file *filp, unsigned long arg) | ||
307 | { | ||
308 | struct f_owner_ex * __user owner_p = (void * __user)arg; | ||
309 | struct f_owner_ex owner; | ||
310 | int ret = 0; | ||
311 | |||
312 | read_lock(&filp->f_owner.lock); | ||
313 | owner.pid = pid_vnr(filp->f_owner.pid); | ||
314 | switch (filp->f_owner.pid_type) { | ||
315 | case PIDTYPE_MAX: | ||
316 | owner.type = F_OWNER_TID; | ||
317 | break; | ||
318 | |||
319 | case PIDTYPE_PID: | ||
320 | owner.type = F_OWNER_PID; | ||
321 | break; | ||
322 | |||
323 | case PIDTYPE_PGID: | ||
324 | owner.type = F_OWNER_GID; | ||
325 | break; | ||
326 | |||
327 | default: | ||
328 | WARN_ON(1); | ||
329 | ret = -EINVAL; | ||
330 | break; | ||
331 | } | ||
332 | read_unlock(&filp->f_owner.lock); | ||
333 | |||
334 | if (!ret) | ||
335 | ret = copy_to_user(owner_p, &owner, sizeof(owner)); | ||
336 | return ret; | ||
337 | } | ||
338 | |||
266 | static long do_fcntl(int fd, unsigned int cmd, unsigned long arg, | 339 | static long do_fcntl(int fd, unsigned int cmd, unsigned long arg, |
267 | struct file *filp) | 340 | struct file *filp) |
268 | { | 341 | { |
@@ -313,6 +386,12 @@ static long do_fcntl(int fd, unsigned int cmd, unsigned long arg, | |||
313 | case F_SETOWN: | 386 | case F_SETOWN: |
314 | err = f_setown(filp, arg, 1); | 387 | err = f_setown(filp, arg, 1); |
315 | break; | 388 | break; |
389 | case F_GETOWN_EX: | ||
390 | err = f_getown_ex(filp, arg); | ||
391 | break; | ||
392 | case F_SETOWN_EX: | ||
393 | err = f_setown_ex(filp, arg); | ||
394 | break; | ||
316 | case F_GETSIG: | 395 | case F_GETSIG: |
317 | err = filp->f_owner.signum; | 396 | err = filp->f_owner.signum; |
318 | break; | 397 | break; |
@@ -428,8 +507,7 @@ static inline int sigio_perm(struct task_struct *p, | |||
428 | 507 | ||
429 | static void send_sigio_to_task(struct task_struct *p, | 508 | static void send_sigio_to_task(struct task_struct *p, |
430 | struct fown_struct *fown, | 509 | struct fown_struct *fown, |
431 | int fd, | 510 | int fd, int reason, int group) |
432 | int reason) | ||
433 | { | 511 | { |
434 | /* | 512 | /* |
435 | * F_SETSIG can change ->signum lockless in parallel, make | 513 | * F_SETSIG can change ->signum lockless in parallel, make |
@@ -461,11 +539,11 @@ static void send_sigio_to_task(struct task_struct *p, | |||
461 | else | 539 | else |
462 | si.si_band = band_table[reason - POLL_IN]; | 540 | si.si_band = band_table[reason - POLL_IN]; |
463 | si.si_fd = fd; | 541 | si.si_fd = fd; |
464 | if (!do_send_sig_info(signum, &si, p, true)) | 542 | if (!do_send_sig_info(signum, &si, p, group)) |
465 | break; | 543 | break; |
466 | /* fall-through: fall back on the old plain SIGIO signal */ | 544 | /* fall-through: fall back on the old plain SIGIO signal */ |
467 | case 0: | 545 | case 0: |
468 | do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, true); | 546 | do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, group); |
469 | } | 547 | } |
470 | } | 548 | } |
471 | 549 | ||
@@ -474,16 +552,23 @@ void send_sigio(struct fown_struct *fown, int fd, int band) | |||
474 | struct task_struct *p; | 552 | struct task_struct *p; |
475 | enum pid_type type; | 553 | enum pid_type type; |
476 | struct pid *pid; | 554 | struct pid *pid; |
555 | int group = 1; | ||
477 | 556 | ||
478 | read_lock(&fown->lock); | 557 | read_lock(&fown->lock); |
558 | |||
479 | type = fown->pid_type; | 559 | type = fown->pid_type; |
560 | if (type == PIDTYPE_MAX) { | ||
561 | group = 0; | ||
562 | type = PIDTYPE_PID; | ||
563 | } | ||
564 | |||
480 | pid = fown->pid; | 565 | pid = fown->pid; |
481 | if (!pid) | 566 | if (!pid) |
482 | goto out_unlock_fown; | 567 | goto out_unlock_fown; |
483 | 568 | ||
484 | read_lock(&tasklist_lock); | 569 | read_lock(&tasklist_lock); |
485 | do_each_pid_task(pid, type, p) { | 570 | do_each_pid_task(pid, type, p) { |
486 | send_sigio_to_task(p, fown, fd, band); | 571 | send_sigio_to_task(p, fown, fd, band, group); |
487 | } while_each_pid_task(pid, type, p); | 572 | } while_each_pid_task(pid, type, p); |
488 | read_unlock(&tasklist_lock); | 573 | read_unlock(&tasklist_lock); |
489 | out_unlock_fown: | 574 | out_unlock_fown: |
@@ -491,10 +576,10 @@ void send_sigio(struct fown_struct *fown, int fd, int band) | |||
491 | } | 576 | } |
492 | 577 | ||
493 | static void send_sigurg_to_task(struct task_struct *p, | 578 | static void send_sigurg_to_task(struct task_struct *p, |
494 | struct fown_struct *fown) | 579 | struct fown_struct *fown, int group) |
495 | { | 580 | { |
496 | if (sigio_perm(p, fown, SIGURG)) | 581 | if (sigio_perm(p, fown, SIGURG)) |
497 | group_send_sig_info(SIGURG, SEND_SIG_PRIV, p); | 582 | do_send_sig_info(SIGURG, SEND_SIG_PRIV, p, group); |
498 | } | 583 | } |
499 | 584 | ||
500 | int send_sigurg(struct fown_struct *fown) | 585 | int send_sigurg(struct fown_struct *fown) |
@@ -502,10 +587,17 @@ int send_sigurg(struct fown_struct *fown) | |||
502 | struct task_struct *p; | 587 | struct task_struct *p; |
503 | enum pid_type type; | 588 | enum pid_type type; |
504 | struct pid *pid; | 589 | struct pid *pid; |
590 | int group = 1; | ||
505 | int ret = 0; | 591 | int ret = 0; |
506 | 592 | ||
507 | read_lock(&fown->lock); | 593 | read_lock(&fown->lock); |
594 | |||
508 | type = fown->pid_type; | 595 | type = fown->pid_type; |
596 | if (type == PIDTYPE_MAX) { | ||
597 | group = 0; | ||
598 | type = PIDTYPE_PID; | ||
599 | } | ||
600 | |||
509 | pid = fown->pid; | 601 | pid = fown->pid; |
510 | if (!pid) | 602 | if (!pid) |
511 | goto out_unlock_fown; | 603 | goto out_unlock_fown; |
@@ -514,7 +606,7 @@ int send_sigurg(struct fown_struct *fown) | |||
514 | 606 | ||
515 | read_lock(&tasklist_lock); | 607 | read_lock(&tasklist_lock); |
516 | do_each_pid_task(pid, type, p) { | 608 | do_each_pid_task(pid, type, p) { |
517 | send_sigurg_to_task(p, fown); | 609 | send_sigurg_to_task(p, fown, group); |
518 | } while_each_pid_task(pid, type, p); | 610 | } while_each_pid_task(pid, type, p); |
519 | read_unlock(&tasklist_lock); | 611 | read_unlock(&tasklist_lock); |
520 | out_unlock_fown: | 612 | out_unlock_fown: |