hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/release-management/issues/IC9Q31 -------------------------------- Add cache mode for fd wakeup in epoll_pwait. In epoll_pwait, read data from ready fd into pre-allocated kernel cache buffer. And then, in sys_read, read from cache buffer and copy to user. So, we can async prefetch read data in epoll_pwait. (lock state can not kernel_read()) +--------------CANCEL<------------------------------+ | ^ | | | | | (prefetch not | (kernel_read() done) | (cache buf | start) | (prefetching) | not empty) +--------------> NONE ---->PREFETCH ----> READY <---+ (vfs_read done or not need) Signed-off-by: Yipeng Zou <zouyipeng@huawei.com> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com> --- arch/Kconfig | 16 +++ fs/eventpoll.c | 255 +++++++++++++++++++++++++++++++++++++++++++++ fs/open.c | 2 + fs/read_write.c | 8 ++ include/linux/fs.h | 33 ++++++ 5 files changed, 314 insertions(+) diff --git a/arch/Kconfig b/arch/Kconfig index 6dc501a4afb1..9a07425c0e1a 100644 --- a/arch/Kconfig +++ b/arch/Kconfig @@ -1205,6 +1205,22 @@ config FAST_SYSCALL exception handling path that only considers necessary features such as security, context saving, and recovery. +config XCALL_PREFETCH + bool "Xcall prefetch support" + select FAST_SYSCALL + default n + help + This enable xcall prefetch feature. + Xcall prefetch feature implements customized epoll_wait() and + read() system calls, which enable data prefetching. + In high-concurrency connection scenarios, this improves + the parallel execution efficiency of the read() system call + and increases the system's business throughput. + The Xcall prefetch feature is suitable for business scenarios + where the epoll I/O multiplexing mechanism is used, the read() + system call takes up a large proportion of time, and the number + of concurrent connections is large. + config ARCH_SUPPORTS_FAST_IRQ bool diff --git a/fs/eventpoll.c b/fs/eventpoll.c index 5ce1ea1f452b..5a1b509e31a6 100644 --- a/fs/eventpoll.c +++ b/fs/eventpoll.c @@ -768,6 +768,63 @@ static void epi_rcu_free(struct rcu_head *head) kmem_cache_free(epi_cache, epi); } +#ifdef CONFIG_XCALL_PREFETCH +#define PREFETCH_ITEM_HASH_BITS 6 +#define PREFETCH_ITEM_TABLE_SIZE (1 << PREFETCH_ITEM_HASH_BITS) +static DEFINE_HASHTABLE(xcall_item_table, PREFETCH_ITEM_HASH_BITS); +static DEFINE_RWLOCK(xcall_table_lock); +static struct workqueue_struct *rc_work; + +static struct prefetch_item *find_prefetch_item(struct file *file) +{ + struct prefetch_item *found = NULL; + unsigned int hash = 0; + + hash = hash_64((u64)file, PREFETCH_ITEM_HASH_BITS); + read_lock(&xcall_table_lock); + hash_for_each_possible(xcall_item_table, found, node, hash) { + if (found->file == file) + break; + } + read_unlock(&xcall_table_lock); + + return found; +} + +void free_prefetch_item(struct file *file) +{ + struct prefetch_item *pfi = find_prefetch_item(file); + + if (!pfi) + return; + + write_lock(&xcall_table_lock); + if (!hlist_unhashed(&pfi->node)) + hlist_del_init(&pfi->node); + write_unlock(&xcall_table_lock); + if (pfi->cache_pages) { + __free_pages(pfi->cache_pages, 0); + pfi->cache = NULL; + } + kfree(pfi); +} + +static void xcall_cancel_work(struct file *file) +{ + struct prefetch_item *pfi; + + if (!current->xcall_select || + !test_bit(__NR_epoll_pwait, current->xcall_select)) + return; + + pfi = find_prefetch_item(file); + if (pfi) + cancel_work_sync(&pfi->work); +} +#else +static inline void xcall_cancel_work(struct file *file) {} +#endif + /* * Removes a "struct epitem" from the eventpoll RB tree and deallocates * all the associated resources. Must be called with "mtx" held. @@ -782,6 +839,7 @@ static int ep_remove(struct eventpoll *ep, struct epitem *epi) * Removes poll wait queue hooks. */ ep_unregister_pollwait(ep, epi); + xcall_cancel_work(file); /* Remove the current item from the list of epoll hooks */ spin_lock(&file->f_lock); @@ -1191,6 +1249,193 @@ static inline bool chain_epi_lockless(struct epitem *epi) return true; } +#ifdef CONFIG_XCALL_PREFETCH +static inline bool transition_state(struct prefetch_item *pfi, + enum cache_state old, enum cache_state new) +{ + return atomic_cmpxchg(&pfi->state, old, new) == old; +} + +static int xcall_read(struct prefetch_item *pfi, unsigned int fd, + char __user *buf, size_t count) +{ + ssize_t copy_ret = -1; + ssize_t copy_len = 0; + + while (!transition_state(pfi, XCALL_CACHE_READY, XCALL_CACHE_CANCEL)) { + if (transition_state(pfi, XCALL_CACHE_NONE, XCALL_CACHE_CANCEL)) + goto reset_pfi_and_retry_vfs_read; + } + + copy_len = pfi->len; + if (unlikely(copy_len < 0)) + goto reset_pfi_and_retry_vfs_read; + + if (copy_len == 0) { + copy_ret = 0; + transition_state(pfi, XCALL_CACHE_CANCEL, XCALL_CACHE_NONE); + goto hit_return; + } + + copy_len = (copy_len >= count) ? count : copy_len; + copy_ret = copy_to_user(buf, (void *)(pfi->cache + pfi->pos), copy_len); + pfi->len -= copy_len; + if (pfi->len <= 0) { + pfi->len = 0; + pfi->pos = 0; + transition_state(pfi, XCALL_CACHE_CANCEL, XCALL_CACHE_NONE); + } else if (pfi->len > 0) { + pfi->pos += copy_len; + transition_state(pfi, XCALL_CACHE_CANCEL, XCALL_CACHE_READY); + } +hit_return: + if (copy_ret == 0) + return copy_len; + else + return -EBADF; + +reset_pfi_and_retry_vfs_read: + pfi->len = 0; + pfi->pos = 0; + cancel_work(&pfi->work); + + return -EAGAIN; +} + +int xcall_read_begin(struct file *file, unsigned int fd, char __user *buf, + size_t count) +{ + struct prefetch_item *pfi = NULL; + + if (!current->xcall_select || + !test_bit(__NR_epoll_pwait, current->xcall_select)) + return -EAGAIN; + + if (!file) + return -EAGAIN; + + pfi = find_prefetch_item(file); + if (!pfi) + return -EAGAIN; + + return xcall_read(pfi, fd, buf, count); +} + +void xcall_read_end(struct file *file) +{ + struct prefetch_item *pfi = NULL; + + if (!current->xcall_select || + !test_bit(__NR_epoll_pwait, current->xcall_select)) + return; + + if (!file) + return; + + pfi = find_prefetch_item(file); + if (!pfi) + return; + + transition_state(pfi, XCALL_CACHE_CANCEL, XCALL_CACHE_NONE); +} + +static void prefetch_work_fn(struct work_struct *work) +{ + struct prefetch_item *pfi = container_of(work, struct prefetch_item, work); + + if (!transition_state(pfi, XCALL_CACHE_NONE, XCALL_CACHE_PREFETCH)) + return; + + pfi->len = kernel_read(pfi->file, pfi->cache, + PAGE_SIZE, &pfi->file->f_pos); + transition_state(pfi, XCALL_CACHE_PREFETCH, XCALL_CACHE_READY); +} + +static void set_prefetch_numa_cpu(struct prefetch_item *pfi, int fd) +{ + int cpu = smp_processor_id(); + + cpumask_and(&pfi->related_cpus, cpu_cpu_mask(cpu), cpu_online_mask); + pfi->cpu = cpumask_next(fd % cpumask_weight(&pfi->related_cpus), + &pfi->related_cpus); +} + +static int get_async_prefetch_cpu(struct prefetch_item *pfi) +{ + int cpu; + + if (pfi->cpu != smp_processor_id()) + return pfi->cpu; + + cpu = cpumask_next(pfi->cpu, &pfi->related_cpus); + if (cpu > cpumask_last(&pfi->related_cpus)) + cpu = cpumask_first(&pfi->related_cpus); + pfi->cpu = cpu; + return pfi->cpu; +} + +static struct prefetch_item *alloc_prefetch_item(struct epitem *epi) +{ + struct file *tfile = epi->ffd.file; + struct prefetch_item *pfi; + int fd = epi->ffd.fd; + + pfi = kmalloc(sizeof(struct prefetch_item), GFP_KERNEL); + if (!pfi) + return NULL; + + pfi->cache_pages = alloc_pages(GFP_KERNEL_ACCOUNT | __GFP_ZERO, 0); + if (!pfi->cache_pages) { + kfree(pfi); + return NULL; + } + + pfi->cache = page_address(pfi->cache_pages); + atomic_set(&pfi->state, XCALL_CACHE_NONE); + INIT_WORK(&pfi->work, prefetch_work_fn); + INIT_HLIST_NODE(&pfi->node); + pfi->fd = fd; + pfi->file = tfile; + pfi->len = 0; + pfi->pos = 0; + set_prefetch_numa_cpu(pfi, fd); + + write_lock(&xcall_table_lock); + hash_add(xcall_item_table, &pfi->node, hash_64((u64)tfile, PREFETCH_ITEM_HASH_BITS)); + write_unlock(&xcall_table_lock); + + return pfi; +} + +static void ep_prefetch_item_enqueue(struct eventpoll *ep, struct epitem *epi) +{ + struct prefetch_item *pfi; + int cpu; + + if (!current->xcall_select || + !test_bit(__NR_epoll_pwait, current->xcall_select)) + return; + + pfi = find_prefetch_item(epi->ffd.file); + if (unlikely(!pfi)) { + pfi = alloc_prefetch_item(epi); + if (unlikely(!pfi)) + return; + } + + if (!pfi->cache || !(epi->event.events & EPOLLIN) || + atomic_read(&pfi->state) != XCALL_CACHE_NONE) + return; + + cpu = get_async_prefetch_cpu(pfi); + queue_work_on(cpu, rc_work, &pfi->work); +} +#else +static void ep_prefetch_item_enqueue(struct eventpoll *ep, struct epitem *epi) +{ +} +#endif + /* * This is the callback that is passed to the wait queue wakeup * mechanism. It is called by the stored file descriptors when they @@ -1751,6 +1996,8 @@ static __poll_t ep_send_events_proc(struct eventpoll *ep, struct list_head *head if (!revents) continue; + ep_prefetch_item_enqueue(ep, epi); + if (__put_user(revents, &uevent->events) || __put_user(epi->event.data, &uevent->data)) { list_add(&epi->rdllink, head); @@ -2454,6 +2701,14 @@ static int __init eventpoll_init(void) pwq_cache = kmem_cache_create("eventpoll_pwq", sizeof(struct eppoll_entry), 0, SLAB_PANIC|SLAB_ACCOUNT, NULL); +#ifdef CONFIG_XCALL_PREFETCH + rc_work = alloc_workqueue("eventpoll_rc", 0, 0); + if (!rc_work) + return -ENOMEM; + + hash_init(xcall_item_table); +#endif + return 0; } fs_initcall(eventpoll_init); diff --git a/fs/open.c b/fs/open.c index 96de0d3f1a8b..381d9ec6e52c 100644 --- a/fs/open.c +++ b/fs/open.c @@ -1287,6 +1287,8 @@ int filp_close(struct file *filp, fl_owner_t id) return 0; } + free_prefetch_item(filp); + if (filp->f_op->flush) retval = filp->f_op->flush(filp, id); diff --git a/fs/read_write.c b/fs/read_write.c index da03b3e65cf3..72f75ab23f57 100644 --- a/fs/read_write.c +++ b/fs/read_write.c @@ -622,6 +622,13 @@ ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count) struct fd f = fdget_pos(fd); ssize_t ret = -EBADF; +#ifdef CONFIG_XCALL_PREFETCH + ret = xcall_read_begin(f.file, fd, buf, count); + if (ret != -EAGAIN) { + fdput_pos(f); + return ret; + } +#endif if (f.file) { loff_t pos, *ppos = file_ppos(f.file); if (ppos) { @@ -632,6 +639,7 @@ ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count) if (ret >= 0 && ppos) f.file->f_pos = pos; fdput_pos(f); + xcall_read_end(f.file); } return ret; } diff --git a/include/linux/fs.h b/include/linux/fs.h index a0ea6b64c45d..2b62da0fc3b7 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -3750,4 +3750,37 @@ static inline bool cachefiles_ondemand_is_enabled(void) } #endif +#ifdef CONFIG_XCALL_PREFETCH +enum cache_state { + XCALL_CACHE_NONE = 0, + XCALL_CACHE_QUEUED, + XCALL_CACHE_PREFETCH, + XCALL_CACHE_READY, + XCALL_CACHE_CANCEL +}; + +struct prefetch_item { + struct file *file; + int fd; + struct work_struct work; + int cpu; + cpumask_t related_cpus; + struct page *cache_pages; + char *cache; + ssize_t len; + /* cache state in epoll_wait */ + atomic_t state; + loff_t pos; + struct hlist_node node; +}; + +int xcall_read_begin(struct file *file, unsigned int fd, char __user *buf, + size_t count); +void xcall_read_end(struct file *file); +void free_prefetch_item(struct file *file); +#else +static inline void xcall_read_end(struct file *file) {} +static inline void free_prefetch_item(struct file *file) {} +#endif + #endif /* _LINUX_FS_H */ -- 2.34.1