This code is ancient, and goes back to when we only had a single page for the pipe buffers. The exact history is hidden in the mists of time (ie "before git", and in fact predates the BK repository too). At that long-ago point in time, it actually helped to try to merge big back-and-forth pipe reads and writes, and not limit pipe reads to the single pipe buffer in length just because that was all we had at a time. However, since then we've expanded the pipe buffers to multiple pages, and this logic really doesn't seem to make sense. And a lot of it is somewhat questionable (ie "hmm, the user asked for a non-blocking read, but we see that there's a writer pending, so let's wait anyway to get the extra data that the writer will have"). But more importantly, it makes the "go to sleep" logic much less obvious, and considering the wakeup issues we've had, I want to make for less of those kinds of things.
Unified · Split
| @@ -348,18 +348,11 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to) · 347 unmodified lines | ||
| 348 | 348 | |
| 349 | 349 | if (!pipe->writers) |
| 350 | 350 | break; |
| 351 | − if (!pipe->waiting_writers) { | |
| 352 | − /* syscall merging: Usually we must not sleep | |
| 353 | − * if O_NONBLOCK is set, or if we got some data. | |
| 354 | − * But if a writer sleeps in kernel space, then | |
| 355 | − * we can wait for that data without violating POSIX. | |
| 356 | − */ | |
| 357 | − if (ret) | |
| 358 | − break; | |
| 359 | − if (filp->f_flags & O_NONBLOCK) { | |
| 360 | − ret = -EAGAIN; | |
| 361 | − break; | |
| 362 | − } | |
| 351 | + if (ret) | |
| 352 | + break; | |
| 353 | + if (filp->f_flags & O_NONBLOCK) { | |
| 354 | + ret = -EAGAIN; | |
| 355 | + break; | |
| 363 | 356 | } |
| 364 | 357 | if (signal_pending(current)) { |
| 365 | 358 | if (!ret) |
| @@ -540,9 +533,7 @@ pipe_write(struct kiocb *iocb, struct iov_iter *from) · 174 unmodified lines | ||
| 540 | 533 | wake_up_interruptible_sync_poll(&pipe->wait, EPOLLIN | EPOLLRDNORM); |
| 541 | 534 | kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN); |
| 542 | 535 | } |
| 543 | − pipe->waiting_writers++; | |
| 544 | 536 | pipe_wait(pipe); |
| 545 | − pipe->waiting_writers--; | |
| 546 | 537 | |
| 547 | 538 | was_empty = pipe_empty(head, pipe->tail); |
| 548 | 539 | } |
| @@ -559,7 +559,7 @@ static int splice_from_pipe_next(struct pipe_inode_info *pipe, struct splice_des · 558 unmodified lines | ||
| 559 | 559 | if (!pipe->writers) |
| 560 | 560 | return 0; |
| 561 | 561 | |
| 562 | − if (!pipe->waiting_writers && sd->num_spliced) | |
| 562 | + if (sd->num_spliced) | |
| 563 | 563 | return 0; |
| 564 | 564 | |
| 565 | 565 | if (sd->flags & SPLICE_F_NONBLOCK) |
| @@ -1098,9 +1098,7 @@ static int wait_for_space(struct pipe_inode_info *pipe, unsigned flags) · 532 unmodified lines | ||
| 1098 | 1098 | return -EAGAIN; |
| 1099 | 1099 | if (signal_pending(current)) |
| 1100 | 1100 | return -ERESTARTSYS; |
| 1101 | − pipe->waiting_writers++; | |
| 1102 | 1101 | pipe_wait(pipe); |
| 1103 | − pipe->waiting_writers--; | |
| 1104 | 1102 | } |
| 1105 | 1103 | } |
| 1106 | 1104 | |
| @@ -1482,11 +1480,9 @@ static int ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags) · 375 unmodified lines | ||
| 1482 | 1480 | } |
| 1483 | 1481 | if (!pipe->writers) |
| 1484 | 1482 | break; |
| 1485 | − if (!pipe->waiting_writers) { | |
| 1486 | − if (flags & SPLICE_F_NONBLOCK) { | |
| 1487 | − ret = -EAGAIN; | |
| 1488 | − break; | |
| 1489 | − } | |
| 1483 | + if (flags & SPLICE_F_NONBLOCK) { | |
| 1484 | + ret = -EAGAIN; | |
| 1485 | + break; | |
| 1490 | 1486 | } |
| 1491 | 1487 | pipe_wait(pipe); |
| 1492 | 1488 | } |
| @@ -1527,9 +1523,7 @@ static int opipe_prep(struct pipe_inode_info *pipe, unsigned int flags) · 34 unmodified lines | ||
| 1527 | 1523 | ret = -ERESTARTSYS; |
| 1528 | 1524 | break; |
| 1529 | 1525 | } |
| 1530 | − pipe->waiting_writers++; | |
| 1531 | 1526 | pipe_wait(pipe); |
| 1532 | − pipe->waiting_writers--; | |
| 1533 | 1527 | } |
| 1534 | 1528 | |
| 1535 | 1529 | pipe_unlock(pipe); |
| @@ -1751,13 +1745,6 @@ static int link_pipe(struct pipe_inode_info *ipipe, · 215 unmodified lines | ||
| 1751 | 1745 | i_tail++; |
| 1752 | 1746 | } while (len); |
| 1753 | 1747 | |
| 1754 | − /* | |
| 1755 | − * return EAGAIN if we have the potential of some data in the | |
| 1756 | − * future, otherwise just return 0 | |
| 1757 | − */ | |
| 1758 | − if (!ret && ipipe->waiting_writers && (flags & SPLICE_F_NONBLOCK)) | |
| 1759 | − ret = -EAGAIN; | |
| 1760 | − | |
| 1761 | 1748 | pipe_unlock(ipipe); |
| 1762 | 1749 | pipe_unlock(opipe); |
| 1763 | 1750 | |
| @@ -38,7 +38,6 @@ struct pipe_buffer { · 37 unmodified lines | ||
| 38 | 38 | * @readers: number of current readers of this pipe |
| 39 | 39 | * @writers: number of current writers of this pipe |
| 40 | 40 | * @files: number of struct file referring this pipe (protected by ->i_lock) |
| 41 | − * @waiting_writers: number of writers blocked waiting for room | |
| 42 | 41 | * @r_counter: reader counter |
| 43 | 42 | * @w_counter: writer counter |
| 44 | 43 | * @fasync_readers: reader side fasync |
| @@ -56,7 +55,6 @@ struct pipe_inode_info { · 11 unmodified lines | ||
| 56 | 55 | unsigned int readers; |
| 57 | 56 | unsigned int writers; |
| 58 | 57 | unsigned int files; |
| 59 | − unsigned int waiting_writers; | |
| 60 | 58 | unsigned int r_counter; |
| 61 | 59 | unsigned int w_counter; |
| 62 | 60 | struct page *tmp_page; |