linux_dsm_epyc7002/include/linux/pipe_fs_i.h
Linus Torvalds 0ddad21d3e pipe: use exclusive waits when reading or writing
This makes the pipe code use separate wait-queues and exclusive waiting
for readers and writers, avoiding a nasty thundering herd problem when
there are lots of readers waiting for data on a pipe (or, less commonly,
lots of writers waiting for a pipe to have space).

While this isn't a common occurrence in the traditional "use a pipe as a
data transport" case, where you typically only have a single reader and
a single writer process, there is one common special case: using a pipe
as a source of "locking tokens" rather than for data communication.

In particular, the GNU make jobserver code ends up using a pipe as a way
to limit parallelism, where each job consumes a token by reading a byte
from the jobserver pipe, and releases the token by writing a byte back
to the pipe.

This pattern is fairly traditional on Unix, and works very well, but
will waste a lot of time waking up a lot of processes when only a single
reader needs to be woken up when a writer releases a new token.

A simplified test-case of just this pipe interaction is to create 64
processes, and then pass a single token around between them (this
test-case also intentionally passes another token that gets ignored to
test the "wake up next" logic too, in case anybody wonders about it):

    #include <unistd.h>

    int main(int argc, char **argv)
    {
        int fd[2], counters[2];

        pipe(fd);
        counters[0] = 0;
        counters[1] = -1;
        write(fd[1], counters, sizeof(counters));

        /* 64 processes */
        fork(); fork(); fork(); fork(); fork(); fork();

        do {
                int i;
                read(fd[0], &i, sizeof(i));
                if (i < 0)
                        continue;
                counters[0] = i+1;
                write(fd[1], counters, (1+(i & 1)) *sizeof(int));
        } while (counters[0] < 1000000);
        return 0;
    }

and in a perfect world, passing that token around should only cause one
context switch per transfer, when the writer of a token causes a
directed wakeup of just a single reader.

But with the "writer wakes all readers" model we traditionally had, on
my test box the above case causes more than an order of magnitude more
scheduling: instead of the expected ~1M context switches, "perf stat"
shows

        231,852.37 msec task-clock                #   15.857 CPUs utilized
        11,250,961      context-switches          #    0.049 M/sec
           616,304      cpu-migrations            #    0.003 M/sec
             1,648      page-faults               #    0.007 K/sec
 1,097,903,998,514      cycles                    #    4.735 GHz
   120,781,778,352      instructions              #    0.11  insn per cycle
    27,997,056,043      branches                  #  120.754 M/sec
       283,581,233      branch-misses             #    1.01% of all branches

      14.621273891 seconds time elapsed

       0.018243000 seconds user
       3.611468000 seconds sys

before this commit.

After this commit, I get

          5,229.55 msec task-clock                #    3.072 CPUs utilized
         1,212,233      context-switches          #    0.232 M/sec
           103,951      cpu-migrations            #    0.020 M/sec
             1,328      page-faults               #    0.254 K/sec
    21,307,456,166      cycles                    #    4.074 GHz
    12,947,819,999      instructions              #    0.61  insn per cycle
     2,881,985,678      branches                  #  551.096 M/sec
        64,267,015      branch-misses             #    2.23% of all branches

       1.702148350 seconds time elapsed

       0.004868000 seconds user
       0.110786000 seconds sys

instead. Much better.

[ Note! This kernel improvement seems to be very good at triggering a
  race condition in the make jobserver (in GNU make 4.2.1) for me. It's
  a long known bug that was fixed back in June 2017 by GNU make commit
  b552b0525198 ("[SV 51159] Use a non-blocking read with pselect to
  avoid hangs.").

  But there wasn't a new release of GNU make until 4.3 on Jan 19 2020,
  so a number of distributions may still have the buggy version. Some
  have backported the fix to their 4.2.1 release, though, and even
  without the fix it's quite timing-dependent whether the bug actually
  is hit. ]

Josh Triplett says:
 "I've been hammering on your pipe fix patch (switching to exclusive
  wait queues) for a month or so, on several different systems, and I've
  run into no issues with it. The patch *substantially* improves
  parallel build times on large (~100 CPU) systems, both with parallel
  make and with other things that use make's pipe-based jobserver.

  All current distributions (including stable and long-term stable
  distributions) have versions of GNU make that no longer have the
  jobserver bug"

Tested-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-08 11:39:19 -08:00

247 lines
7.5 KiB
C

/* SPDX-License-Identifier: GPL-2.0 */
#ifndef _LINUX_PIPE_FS_I_H
#define _LINUX_PIPE_FS_I_H
#define PIPE_DEF_BUFFERS 16
#define PIPE_BUF_FLAG_LRU 0x01 /* page is on the LRU */
#define PIPE_BUF_FLAG_ATOMIC 0x02 /* was atomically mapped */
#define PIPE_BUF_FLAG_GIFT 0x04 /* page is a gift */
#define PIPE_BUF_FLAG_PACKET 0x08 /* read() as a packet */
/**
* struct pipe_buffer - a linux kernel pipe buffer
* @page: the page containing the data for the pipe buffer
* @offset: offset of data inside the @page
* @len: length of data inside the @page
* @ops: operations associated with this buffer. See @pipe_buf_operations.
* @flags: pipe buffer flags. See above.
* @private: private data owned by the ops.
**/
struct pipe_buffer {
struct page *page;
unsigned int offset, len;
const struct pipe_buf_operations *ops;
unsigned int flags;
unsigned long private;
};
/**
* struct pipe_inode_info - a linux kernel pipe
* @mutex: mutex protecting the whole thing
* @wait: reader/writer wait point in case of empty/full pipe
* @head: The point of buffer production
* @tail: The point of buffer consumption
* @max_usage: The maximum number of slots that may be used in the ring
* @ring_size: total number of buffers (should be a power of 2)
* @tmp_page: cached released page
* @readers: number of current readers of this pipe
* @writers: number of current writers of this pipe
* @files: number of struct file referring this pipe (protected by ->i_lock)
* @r_counter: reader counter
* @w_counter: writer counter
* @fasync_readers: reader side fasync
* @fasync_writers: writer side fasync
* @bufs: the circular array of pipe buffers
* @user: the user who created this pipe
**/
struct pipe_inode_info {
struct mutex mutex;
wait_queue_head_t rd_wait, wr_wait;
unsigned int head;
unsigned int tail;
unsigned int max_usage;
unsigned int ring_size;
unsigned int readers;
unsigned int writers;
unsigned int files;
unsigned int r_counter;
unsigned int w_counter;
struct page *tmp_page;
struct fasync_struct *fasync_readers;
struct fasync_struct *fasync_writers;
struct pipe_buffer *bufs;
struct user_struct *user;
};
/*
* Note on the nesting of these functions:
*
* ->confirm()
* ->steal()
*
* That is, ->steal() must be called on a confirmed buffer.
* See below for the meaning of each operation. Also see kerneldoc
* in fs/pipe.c for the pipe and generic variants of these hooks.
*/
struct pipe_buf_operations {
/*
* ->confirm() verifies that the data in the pipe buffer is there
* and that the contents are good. If the pages in the pipe belong
* to a file system, we may need to wait for IO completion in this
* hook. Returns 0 for good, or a negative error value in case of
* error.
*/
int (*confirm)(struct pipe_inode_info *, struct pipe_buffer *);
/*
* When the contents of this pipe buffer has been completely
* consumed by a reader, ->release() is called.
*/
void (*release)(struct pipe_inode_info *, struct pipe_buffer *);
/*
* Attempt to take ownership of the pipe buffer and its contents.
* ->steal() returns 0 for success, in which case the contents
* of the pipe (the buf->page) is locked and now completely owned
* by the caller. The page may then be transferred to a different
* mapping, the most often used case is insertion into different
* file address space cache.
*/
int (*steal)(struct pipe_inode_info *, struct pipe_buffer *);
/*
* Get a reference to the pipe buffer.
*/
bool (*get)(struct pipe_inode_info *, struct pipe_buffer *);
};
/**
* pipe_empty - Return true if the pipe is empty
* @head: The pipe ring head pointer
* @tail: The pipe ring tail pointer
*/
static inline bool pipe_empty(unsigned int head, unsigned int tail)
{
return head == tail;
}
/**
* pipe_occupancy - Return number of slots used in the pipe
* @head: The pipe ring head pointer
* @tail: The pipe ring tail pointer
*/
static inline unsigned int pipe_occupancy(unsigned int head, unsigned int tail)
{
return head - tail;
}
/**
* pipe_full - Return true if the pipe is full
* @head: The pipe ring head pointer
* @tail: The pipe ring tail pointer
* @limit: The maximum amount of slots available.
*/
static inline bool pipe_full(unsigned int head, unsigned int tail,
unsigned int limit)
{
return pipe_occupancy(head, tail) >= limit;
}
/**
* pipe_space_for_user - Return number of slots available to userspace
* @head: The pipe ring head pointer
* @tail: The pipe ring tail pointer
* @pipe: The pipe info structure
*/
static inline unsigned int pipe_space_for_user(unsigned int head, unsigned int tail,
struct pipe_inode_info *pipe)
{
unsigned int p_occupancy, p_space;
p_occupancy = pipe_occupancy(head, tail);
if (p_occupancy >= pipe->max_usage)
return 0;
p_space = pipe->ring_size - p_occupancy;
if (p_space > pipe->max_usage)
p_space = pipe->max_usage;
return p_space;
}
/**
* pipe_buf_get - get a reference to a pipe_buffer
* @pipe: the pipe that the buffer belongs to
* @buf: the buffer to get a reference to
*
* Return: %true if the reference was successfully obtained.
*/
static inline __must_check bool pipe_buf_get(struct pipe_inode_info *pipe,
struct pipe_buffer *buf)
{
return buf->ops->get(pipe, buf);
}
/**
* pipe_buf_release - put a reference to a pipe_buffer
* @pipe: the pipe that the buffer belongs to
* @buf: the buffer to put a reference to
*/
static inline void pipe_buf_release(struct pipe_inode_info *pipe,
struct pipe_buffer *buf)
{
const struct pipe_buf_operations *ops = buf->ops;
buf->ops = NULL;
ops->release(pipe, buf);
}
/**
* pipe_buf_confirm - verify contents of the pipe buffer
* @pipe: the pipe that the buffer belongs to
* @buf: the buffer to confirm
*/
static inline int pipe_buf_confirm(struct pipe_inode_info *pipe,
struct pipe_buffer *buf)
{
return buf->ops->confirm(pipe, buf);
}
/**
* pipe_buf_steal - attempt to take ownership of a pipe_buffer
* @pipe: the pipe that the buffer belongs to
* @buf: the buffer to attempt to steal
*/
static inline int pipe_buf_steal(struct pipe_inode_info *pipe,
struct pipe_buffer *buf)
{
return buf->ops->steal(pipe, buf);
}
/* Differs from PIPE_BUF in that PIPE_SIZE is the length of the actual
memory allocation, whereas PIPE_BUF makes atomicity guarantees. */
#define PIPE_SIZE PAGE_SIZE
/* Pipe lock and unlock operations */
void pipe_lock(struct pipe_inode_info *);
void pipe_unlock(struct pipe_inode_info *);
void pipe_double_lock(struct pipe_inode_info *, struct pipe_inode_info *);
extern unsigned int pipe_max_size;
extern unsigned long pipe_user_pages_hard;
extern unsigned long pipe_user_pages_soft;
/* Drop the inode semaphore and wait for a pipe event, atomically */
void pipe_wait(struct pipe_inode_info *pipe);
struct pipe_inode_info *alloc_pipe_info(void);
void free_pipe_info(struct pipe_inode_info *);
/* Generic pipe buffer ops functions */
bool generic_pipe_buf_get(struct pipe_inode_info *, struct pipe_buffer *);
int generic_pipe_buf_confirm(struct pipe_inode_info *, struct pipe_buffer *);
int generic_pipe_buf_steal(struct pipe_inode_info *, struct pipe_buffer *);
int generic_pipe_buf_nosteal(struct pipe_inode_info *, struct pipe_buffer *);
void generic_pipe_buf_release(struct pipe_inode_info *, struct pipe_buffer *);
void pipe_buf_mark_unmergeable(struct pipe_buffer *buf);
extern const struct pipe_buf_operations nosteal_pipe_buf_ops;
/* for F_SETPIPE_SZ and F_GETPIPE_SZ */
long pipe_fcntl(struct file *, unsigned int, unsigned long arg);
struct pipe_inode_info *get_pipe_info(struct file *file);
int create_pipe_files(struct file **, int);
unsigned int round_pipe_size(unsigned long size);
#endif