From 144cba1493fdd6e3e1980e439a31df877831ebcd Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Mon, 27 Apr 2015 11:09:54 +0800 Subject: libceph: allow setting osd_req_op's flags Signed-off-by: Yan, Zheng Reviewed-by: Alex Elder --- include/linux/ceph/osd_client.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include') diff --git a/include/linux/ceph/osd_client.h b/include/linux/ceph/osd_client.h index 61b19c46bdb3..7506b485bb6d 100644 --- a/include/linux/ceph/osd_client.h +++ b/include/linux/ceph/osd_client.h @@ -249,7 +249,7 @@ extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg); extern void osd_req_op_init(struct ceph_osd_request *osd_req, - unsigned int which, u16 opcode); + unsigned int which, u16 opcode, u32 flags); extern void osd_req_op_raw_data_in_pages(struct ceph_osd_request *, unsigned int which, -- cgit v1.2.3 From d50c97b566c5bbf990eff472e9feaa58fdebdd33 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Fri, 15 May 2015 11:52:20 +0300 Subject: libceph: nuke time_sub() Unused since ceph got merged into mainline I guess. Signed-off-by: Ilya Dryomov Reviewed-by: Alex Elder --- include/linux/ceph/libceph.h | 9 --------- 1 file changed, 9 deletions(-) (limited to 'include') diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h index 30f92cefaa72..85ae9a889a3f 100644 --- a/include/linux/ceph/libceph.h +++ b/include/linux/ceph/libceph.h @@ -93,15 +93,6 @@ enum { CEPH_MOUNT_SHUTDOWN, }; -/* - * subtract jiffies - */ -static inline unsigned long time_sub(unsigned long a, unsigned long b) -{ - BUG_ON(time_after(b, a)); - return (long)a - (long)b; -} - struct ceph_mds_client; /* -- cgit v1.2.3 From a319bf56a617354e62cf5f774d2ca4e1a8a3bff3 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Fri, 15 May 2015 12:02:17 +0300 Subject: libceph: store timeouts in jiffies, verify user input There are currently three libceph-level timeouts that the user can specify on mount: mount_timeout, osd_idle_ttl and osdkeepalive. All of these are in seconds and no checking is done on user input: negative values are accepted, we multiply them all by HZ which may or may not overflow, arbitrarily large jiffies then get added together, etc. There is also a bug in the way mount_timeout=0 is handled. It's supposed to mean "infinite timeout", but that's not how wait.h APIs treat it and so __ceph_open_session() for example will busy loop without much chance of being interrupted if none of ceph-mons are there. Fix all this by verifying user input, storing timeouts capped by msecs_to_jiffies() in jiffies and using the new ceph_timeout_jiffies() helper for all user-specified waits to handle infinite timeouts correctly. Signed-off-by: Ilya Dryomov Reviewed-by: Alex Elder --- drivers/block/rbd.c | 5 +++-- fs/ceph/dir.c | 4 ++-- fs/ceph/mds_client.c | 12 ++++++------ fs/ceph/mds_client.h | 2 +- fs/ceph/super.c | 2 +- include/linux/ceph/libceph.h | 17 +++++++++++------ net/ceph/ceph_common.c | 41 +++++++++++++++++++++++++++++++---------- net/ceph/mon_client.c | 11 +++++++++-- net/ceph/osd_client.c | 15 +++++++-------- 9 files changed, 71 insertions(+), 38 deletions(-) (limited to 'include') diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 349115ae3bc2..992683b6b299 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -4963,8 +4963,8 @@ out_err: */ static int rbd_add_get_pool_id(struct rbd_client *rbdc, const char *pool_name) { + struct ceph_options *opts = rbdc->client->options; u64 newest_epoch; - unsigned long timeout = rbdc->client->options->mount_timeout * HZ; int tries = 0; int ret; @@ -4979,7 +4979,8 @@ again: if (rbdc->client->osdc.osdmap->epoch < newest_epoch) { ceph_monc_request_next_osdmap(&rbdc->client->monc); (void) ceph_monc_wait_osdmap(&rbdc->client->monc, - newest_epoch, timeout); + newest_epoch, + opts->mount_timeout); goto again; } else { /* the osdmap we have is new enough */ diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c index 4248307fea90..173dd4b58c71 100644 --- a/fs/ceph/dir.c +++ b/fs/ceph/dir.c @@ -1259,8 +1259,8 @@ static int ceph_dir_fsync(struct file *file, loff_t start, loff_t end, inode, req->r_tid, last_tid); if (req->r_timeout) { unsigned long time_left = wait_for_completion_timeout( - &req->r_safe_completion, - req->r_timeout); + &req->r_safe_completion, + ceph_timeout_jiffies(req->r_timeout)); if (time_left > 0) ret = 0; else diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c index 69a36f40517f..0b0e0a9a81c0 100644 --- a/fs/ceph/mds_client.c +++ b/fs/ceph/mds_client.c @@ -2268,7 +2268,8 @@ int ceph_mdsc_do_request(struct ceph_mds_client *mdsc, dout("do_request waiting\n"); if (req->r_timeout) { err = (long)wait_for_completion_killable_timeout( - &req->r_completion, req->r_timeout); + &req->r_completion, + ceph_timeout_jiffies(req->r_timeout)); if (err == 0) err = -EIO; } else if (req->r_wait_for_completion) { @@ -3424,8 +3425,8 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) */ static void wait_requests(struct ceph_mds_client *mdsc) { + struct ceph_options *opts = mdsc->fsc->client->options; struct ceph_mds_request *req; - struct ceph_fs_client *fsc = mdsc->fsc; mutex_lock(&mdsc->mutex); if (__get_oldest_req(mdsc)) { @@ -3433,7 +3434,7 @@ static void wait_requests(struct ceph_mds_client *mdsc) dout("wait_requests waiting for requests\n"); wait_for_completion_timeout(&mdsc->safe_umount_waiters, - fsc->client->options->mount_timeout * HZ); + ceph_timeout_jiffies(opts->mount_timeout)); /* tear down remaining requests */ mutex_lock(&mdsc->mutex); @@ -3556,10 +3557,9 @@ static bool done_closing_sessions(struct ceph_mds_client *mdsc) */ void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc) { + struct ceph_options *opts = mdsc->fsc->client->options; struct ceph_mds_session *session; int i; - struct ceph_fs_client *fsc = mdsc->fsc; - unsigned long timeout = fsc->client->options->mount_timeout * HZ; dout("close_sessions\n"); @@ -3580,7 +3580,7 @@ void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc) dout("waiting for sessions to close\n"); wait_event_timeout(mdsc->session_close_wq, done_closing_sessions(mdsc), - timeout); + ceph_timeout_jiffies(opts->mount_timeout)); /* tear down remaining sessions */ mutex_lock(&mdsc->mutex); diff --git a/fs/ceph/mds_client.h b/fs/ceph/mds_client.h index 2ef799961ebb..509d6822e9b1 100644 --- a/fs/ceph/mds_client.h +++ b/fs/ceph/mds_client.h @@ -227,7 +227,7 @@ struct ceph_mds_request { int r_err; bool r_aborted; - unsigned long r_timeout; /* optional. jiffies */ + unsigned long r_timeout; /* optional. jiffies, 0 is "wait forever" */ unsigned long r_started; /* start time to measure timeout against */ unsigned long r_request_started; /* start time for mds request only, used to measure lease durations */ diff --git a/fs/ceph/super.c b/fs/ceph/super.c index 9a5350030af8..edeb83c43112 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -742,7 +742,7 @@ static struct dentry *open_root_dentry(struct ceph_fs_client *fsc, req->r_ino1.ino = CEPH_INO_ROOT; req->r_ino1.snap = CEPH_NOSNAP; req->r_started = started; - req->r_timeout = fsc->client->options->mount_timeout * HZ; + req->r_timeout = fsc->client->options->mount_timeout; req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE); req->r_num_caps = 2; err = ceph_mdsc_do_request(mdsc, NULL, req); diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h index 85ae9a889a3f..d73a569f9bf5 100644 --- a/include/linux/ceph/libceph.h +++ b/include/linux/ceph/libceph.h @@ -43,9 +43,9 @@ struct ceph_options { int flags; struct ceph_fsid fsid; struct ceph_entity_addr my_addr; - int mount_timeout; - int osd_idle_ttl; - int osd_keepalive_timeout; + unsigned long mount_timeout; /* jiffies */ + unsigned long osd_idle_ttl; /* jiffies */ + unsigned long osd_keepalive_timeout; /* jiffies */ /* * any type that can't be simply compared or doesn't need need @@ -63,9 +63,9 @@ struct ceph_options { /* * defaults */ -#define CEPH_MOUNT_TIMEOUT_DEFAULT 60 -#define CEPH_OSD_KEEPALIVE_DEFAULT 5 -#define CEPH_OSD_IDLE_TTL_DEFAULT 60 +#define CEPH_MOUNT_TIMEOUT_DEFAULT msecs_to_jiffies(60 * 1000) +#define CEPH_OSD_KEEPALIVE_DEFAULT msecs_to_jiffies(5 * 1000) +#define CEPH_OSD_IDLE_TTL_DEFAULT msecs_to_jiffies(60 * 1000) #define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024) #define CEPH_MSG_MAX_MIDDLE_LEN (16*1024*1024) @@ -93,6 +93,11 @@ enum { CEPH_MOUNT_SHUTDOWN, }; +static inline unsigned long ceph_timeout_jiffies(unsigned long timeout) +{ + return timeout ?: MAX_SCHEDULE_TIMEOUT; +} + struct ceph_mds_client; /* diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c index 79e8f71aef5b..a80e91c2c9a3 100644 --- a/net/ceph/ceph_common.c +++ b/net/ceph/ceph_common.c @@ -352,8 +352,8 @@ ceph_parse_options(char *options, const char *dev_name, /* start with defaults */ opt->flags = CEPH_OPT_DEFAULT; opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT; - opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */ - opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */ + opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; + opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* get mon ip(s) */ /* ip1[:port1][,ip2[:port2]...] */ @@ -439,13 +439,32 @@ ceph_parse_options(char *options, const char *dev_name, pr_warn("ignoring deprecated osdtimeout option\n"); break; case Opt_osdkeepalivetimeout: - opt->osd_keepalive_timeout = intval; + /* 0 isn't well defined right now, reject it */ + if (intval < 1 || intval > INT_MAX / 1000) { + pr_err("osdkeepalive out of range\n"); + err = -EINVAL; + goto out; + } + opt->osd_keepalive_timeout = + msecs_to_jiffies(intval * 1000); break; case Opt_osd_idle_ttl: - opt->osd_idle_ttl = intval; + /* 0 isn't well defined right now, reject it */ + if (intval < 1 || intval > INT_MAX / 1000) { + pr_err("osd_idle_ttl out of range\n"); + err = -EINVAL; + goto out; + } + opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000); break; case Opt_mount_timeout: - opt->mount_timeout = intval; + /* 0 is "wait forever" (i.e. infinite timeout) */ + if (intval < 0 || intval > INT_MAX / 1000) { + pr_err("mount_timeout out of range\n"); + err = -EINVAL; + goto out; + } + opt->mount_timeout = msecs_to_jiffies(intval * 1000); break; case Opt_share: @@ -512,12 +531,14 @@ int ceph_print_client_options(struct seq_file *m, struct ceph_client *client) seq_puts(m, "notcp_nodelay,"); if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT) - seq_printf(m, "mount_timeout=%d,", opt->mount_timeout); + seq_printf(m, "mount_timeout=%d,", + jiffies_to_msecs(opt->mount_timeout) / 1000); if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT) - seq_printf(m, "osd_idle_ttl=%d,", opt->osd_idle_ttl); + seq_printf(m, "osd_idle_ttl=%d,", + jiffies_to_msecs(opt->osd_idle_ttl) / 1000); if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT) seq_printf(m, "osdkeepalivetimeout=%d,", - opt->osd_keepalive_timeout); + jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000); /* drop redundant comma */ if (m->count != pos) @@ -627,7 +648,7 @@ static int have_mon_and_osd_map(struct ceph_client *client) int __ceph_open_session(struct ceph_client *client, unsigned long started) { int err; - unsigned long timeout = client->options->mount_timeout * HZ; + unsigned long timeout = client->options->mount_timeout; /* open session, and wait for mon and osd maps */ err = ceph_monc_open_session(&client->monc); @@ -643,7 +664,7 @@ int __ceph_open_session(struct ceph_client *client, unsigned long started) dout("mount waiting for mon_map\n"); err = wait_event_interruptible_timeout(client->auth_wq, have_mon_and_osd_map(client) || (client->auth_err < 0), - timeout); + ceph_timeout_jiffies(timeout)); if (err == -EINTR || err == -ERESTARTSYS) return err; if (client->auth_err < 0) diff --git a/net/ceph/mon_client.c b/net/ceph/mon_client.c index 2b3cf05e87b0..0da3bdc116f7 100644 --- a/net/ceph/mon_client.c +++ b/net/ceph/mon_client.c @@ -298,6 +298,12 @@ void ceph_monc_request_next_osdmap(struct ceph_mon_client *monc) } EXPORT_SYMBOL(ceph_monc_request_next_osdmap); +/* + * Wait for an osdmap with a given epoch. + * + * @epoch: epoch to wait for + * @timeout: in jiffies, 0 means "wait forever" + */ int ceph_monc_wait_osdmap(struct ceph_mon_client *monc, u32 epoch, unsigned long timeout) { @@ -308,11 +314,12 @@ int ceph_monc_wait_osdmap(struct ceph_mon_client *monc, u32 epoch, while (monc->have_osdmap < epoch) { mutex_unlock(&monc->mutex); - if (timeout != 0 && time_after_eq(jiffies, started + timeout)) + if (timeout && time_after_eq(jiffies, started + timeout)) return -ETIMEDOUT; ret = wait_event_interruptible_timeout(monc->client->auth_wq, - monc->have_osdmap >= epoch, timeout); + monc->have_osdmap >= epoch, + ceph_timeout_jiffies(timeout)); if (ret < 0) return ret; diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index 4cb4fab46e4f..50033677c0fa 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -1097,7 +1097,7 @@ static void __move_osd_to_lru(struct ceph_osd_client *osdc, BUG_ON(!list_empty(&osd->o_osd_lru)); list_add_tail(&osd->o_osd_lru, &osdc->osd_lru); - osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl * HZ; + osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl; } static void maybe_move_osd_to_lru(struct ceph_osd_client *osdc, @@ -1208,7 +1208,7 @@ static struct ceph_osd *__lookup_osd(struct ceph_osd_client *osdc, int o) static void __schedule_osd_timeout(struct ceph_osd_client *osdc) { schedule_delayed_work(&osdc->timeout_work, - osdc->client->options->osd_keepalive_timeout * HZ); + osdc->client->options->osd_keepalive_timeout); } static void __cancel_osd_timeout(struct ceph_osd_client *osdc) @@ -1576,10 +1576,9 @@ static void handle_timeout(struct work_struct *work) { struct ceph_osd_client *osdc = container_of(work, struct ceph_osd_client, timeout_work.work); + struct ceph_options *opts = osdc->client->options; struct ceph_osd_request *req; struct ceph_osd *osd; - unsigned long keepalive = - osdc->client->options->osd_keepalive_timeout * HZ; struct list_head slow_osds; dout("timeout\n"); down_read(&osdc->map_sem); @@ -1595,7 +1594,8 @@ static void handle_timeout(struct work_struct *work) */ INIT_LIST_HEAD(&slow_osds); list_for_each_entry(req, &osdc->req_lru, r_req_lru_item) { - if (time_before(jiffies, req->r_stamp + keepalive)) + if (time_before(jiffies, + req->r_stamp + opts->osd_keepalive_timeout)) break; osd = req->r_osd; @@ -1622,8 +1622,7 @@ static void handle_osds_timeout(struct work_struct *work) struct ceph_osd_client *osdc = container_of(work, struct ceph_osd_client, osds_timeout_work.work); - unsigned long delay = - osdc->client->options->osd_idle_ttl * HZ >> 2; + unsigned long delay = osdc->client->options->osd_idle_ttl / 4; dout("osds timeout\n"); down_read(&osdc->map_sem); @@ -2628,7 +2627,7 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) osdc->event_count = 0; schedule_delayed_work(&osdc->osds_timeout_work, - round_jiffies_relative(osdc->client->options->osd_idle_ttl * HZ)); + round_jiffies_relative(osdc->client->options->osd_idle_ttl)); err = -ENOMEM; osdc->req_mempool = mempool_create_kmalloc_pool(10, -- cgit v1.2.3 From f66fd9f0952187d274c13c136b74548f792c1925 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 10 Jun 2015 17:26:13 +0800 Subject: ceph: pre-allocate data structure that tracks caps flushing Signed-off-by: Yan, Zheng --- fs/ceph/addr.c | 19 +++++++++++++++---- fs/ceph/caps.c | 26 ++++++++++++++++++++++---- fs/ceph/file.c | 18 ++++++++++++++++-- fs/ceph/inode.c | 15 +++++++++++++-- fs/ceph/mds_client.c | 6 +++++- fs/ceph/super.c | 8 ++++++++ fs/ceph/super.h | 6 +++++- fs/ceph/xattr.c | 20 ++++++++++++++++++-- include/linux/ceph/libceph.h | 1 + 9 files changed, 103 insertions(+), 16 deletions(-) (limited to 'include') diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 5f53ac0d9d7c..7edf3c49e661 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -1308,12 +1308,17 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) struct inode *inode = file_inode(vma->vm_file); struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_file_info *fi = vma->vm_file->private_data; + struct ceph_cap_flush *prealloc_cf; struct page *page = vmf->page; loff_t off = page_offset(page); loff_t size = i_size_read(inode); size_t len; int want, got, ret; + prealloc_cf = ceph_alloc_cap_flush(); + if (!prealloc_cf) + return VM_FAULT_SIGBUS; + if (ci->i_inline_version != CEPH_INLINE_NONE) { struct page *locked_page = NULL; if (off == 0) { @@ -1323,8 +1328,10 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) ret = ceph_uninline_data(vma->vm_file, locked_page); if (locked_page) unlock_page(locked_page); - if (ret < 0) - return VM_FAULT_SIGBUS; + if (ret < 0) { + ret = VM_FAULT_SIGBUS; + goto out_free; + } } if (off + PAGE_CACHE_SIZE <= size) @@ -1346,7 +1353,8 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) break; if (ret != -ERESTARTSYS) { WARN_ON(1); - return VM_FAULT_SIGBUS; + ret = VM_FAULT_SIGBUS; + goto out_free; } } dout("page_mkwrite %p %llu~%zd got cap refs on %s\n", @@ -1381,7 +1389,8 @@ out: int dirty; spin_lock(&ci->i_ceph_lock); ci->i_inline_version = CEPH_INLINE_NONE; - dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR); + dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, + &prealloc_cf); spin_unlock(&ci->i_ceph_lock); if (dirty) __mark_inode_dirty(inode, dirty); @@ -1390,6 +1399,8 @@ out: dout("page_mkwrite %p %llu~%zd dropping cap refs on %s ret %d\n", inode, off, len, ceph_cap_string(got), ret); ceph_put_cap_refs(ci, got); +out_free: + ceph_free_cap_flush(prealloc_cf); return ret; } diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 69a16044ec41..dd7b20adf1d4 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -1356,7 +1356,8 @@ static void ceph_flush_snaps(struct ceph_inode_info *ci) * Caller is then responsible for calling __mark_inode_dirty with the * returned flags value. */ -int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask) +int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, + struct ceph_cap_flush **pcf) { struct ceph_mds_client *mdsc = ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc; @@ -1376,6 +1377,9 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask) ceph_cap_string(was | mask)); ci->i_dirty_caps |= mask; if (was == 0) { + WARN_ON_ONCE(ci->i_prealloc_cap_flush); + swap(ci->i_prealloc_cap_flush, *pcf); + if (!ci->i_head_snapc) { WARN_ON_ONCE(!rwsem_is_locked(&mdsc->snap_rwsem)); ci->i_head_snapc = ceph_get_snap_context( @@ -1391,6 +1395,8 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask) ihold(inode); dirty |= I_DIRTY_SYNC; } + } else { + WARN_ON_ONCE(!ci->i_prealloc_cap_flush); } BUG_ON(list_empty(&ci->i_dirty_item)); if (((was | ci->i_flushing_caps) & CEPH_CAP_FILE_BUFFER) && @@ -1446,6 +1452,17 @@ static void __add_cap_flushing_to_mdsc(struct ceph_mds_client *mdsc, rb_insert_color(&cf->g_node, &mdsc->cap_flush_tree); } +struct ceph_cap_flush *ceph_alloc_cap_flush(void) +{ + return kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL); +} + +void ceph_free_cap_flush(struct ceph_cap_flush *cf) +{ + if (cf) + kmem_cache_free(ceph_cap_flush_cachep, cf); +} + static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc) { struct rb_node *n = rb_first(&mdsc->cap_flush_tree); @@ -1469,11 +1486,12 @@ static int __mark_caps_flushing(struct inode *inode, { struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc; struct ceph_inode_info *ci = ceph_inode(inode); - struct ceph_cap_flush *cf; + struct ceph_cap_flush *cf = NULL; int flushing; BUG_ON(ci->i_dirty_caps == 0); BUG_ON(list_empty(&ci->i_dirty_item)); + BUG_ON(!ci->i_prealloc_cap_flush); flushing = ci->i_dirty_caps; dout("__mark_caps_flushing flushing %s, flushing_caps %s -> %s\n", @@ -1484,7 +1502,7 @@ static int __mark_caps_flushing(struct inode *inode, ci->i_dirty_caps = 0; dout(" inode %p now !dirty\n", inode); - cf = kmalloc(sizeof(*cf), GFP_ATOMIC); + swap(cf, ci->i_prealloc_cap_flush); cf->caps = flushing; cf->kick = false; @@ -3075,7 +3093,7 @@ out: cf = list_first_entry(&to_remove, struct ceph_cap_flush, list); list_del(&cf->list); - kfree(cf); + ceph_free_cap_flush(cf); } if (drop) iput(inode); diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 0a76a370d798..8a4eb4d21d3c 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -939,6 +939,7 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_osd_client *osdc = &ceph_sb_to_client(inode->i_sb)->client->osdc; + struct ceph_cap_flush *prealloc_cf; ssize_t count, written = 0; int err, want, got; loff_t pos; @@ -946,6 +947,10 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) if (ceph_snap(inode) != CEPH_NOSNAP) return -EROFS; + prealloc_cf = ceph_alloc_cap_flush(); + if (!prealloc_cf) + return -ENOMEM; + mutex_lock(&inode->i_mutex); /* We can write back this queue in page reclaim */ @@ -1050,7 +1055,8 @@ retry_snap: int dirty; spin_lock(&ci->i_ceph_lock); ci->i_inline_version = CEPH_INLINE_NONE; - dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR); + dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, + &prealloc_cf); spin_unlock(&ci->i_ceph_lock); if (dirty) __mark_inode_dirty(inode, dirty); @@ -1074,6 +1080,7 @@ retry_snap: out: mutex_unlock(&inode->i_mutex); out_unlocked: + ceph_free_cap_flush(prealloc_cf); current->backing_dev_info = NULL; return written ? written : err; } @@ -1270,6 +1277,7 @@ static long ceph_fallocate(struct file *file, int mode, struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_osd_client *osdc = &ceph_inode_to_client(inode)->client->osdc; + struct ceph_cap_flush *prealloc_cf; int want, got = 0; int dirty; int ret = 0; @@ -1282,6 +1290,10 @@ static long ceph_fallocate(struct file *file, int mode, if (!S_ISREG(inode->i_mode)) return -EOPNOTSUPP; + prealloc_cf = ceph_alloc_cap_flush(); + if (!prealloc_cf) + return -ENOMEM; + mutex_lock(&inode->i_mutex); if (ceph_snap(inode) != CEPH_NOSNAP) { @@ -1328,7 +1340,8 @@ static long ceph_fallocate(struct file *file, int mode, if (!ret) { spin_lock(&ci->i_ceph_lock); ci->i_inline_version = CEPH_INLINE_NONE; - dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR); + dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, + &prealloc_cf); spin_unlock(&ci->i_ceph_lock); if (dirty) __mark_inode_dirty(inode, dirty); @@ -1337,6 +1350,7 @@ static long ceph_fallocate(struct file *file, int mode, ceph_put_cap_refs(ci, got); unlock: mutex_unlock(&inode->i_mutex); + ceph_free_cap_flush(prealloc_cf); return ret; } diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 3326302f5884..e86d1a4efc46 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -416,6 +416,7 @@ struct inode *ceph_alloc_inode(struct super_block *sb) ci->i_flushing_caps = 0; INIT_LIST_HEAD(&ci->i_dirty_item); INIT_LIST_HEAD(&ci->i_flushing_item); + ci->i_prealloc_cap_flush = NULL; ci->i_cap_flush_tree = RB_ROOT; init_waitqueue_head(&ci->i_cap_wq); ci->i_hold_caps_min = 0; @@ -1720,6 +1721,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) const unsigned int ia_valid = attr->ia_valid; struct ceph_mds_request *req; struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc; + struct ceph_cap_flush *prealloc_cf; int issued; int release = 0, dirtied = 0; int mask = 0; @@ -1734,10 +1736,16 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) if (err != 0) return err; + prealloc_cf = ceph_alloc_cap_flush(); + if (!prealloc_cf) + return -ENOMEM; + req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETATTR, USE_AUTH_MDS); - if (IS_ERR(req)) + if (IS_ERR(req)) { + ceph_free_cap_flush(prealloc_cf); return PTR_ERR(req); + } spin_lock(&ci->i_ceph_lock); issued = __ceph_caps_issued(ci, NULL); @@ -1895,7 +1903,8 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) dout("setattr %p ATTR_FILE ... hrm!\n", inode); if (dirtied) { - inode_dirty_flags = __ceph_mark_dirty_caps(ci, dirtied); + inode_dirty_flags = __ceph_mark_dirty_caps(ci, dirtied, + &prealloc_cf); inode->i_ctime = CURRENT_TIME; } @@ -1927,9 +1936,11 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) ceph_mdsc_put_request(req); if (mask & CEPH_SETATTR_SIZE) __ceph_do_pending_vmtruncate(inode); + ceph_free_cap_flush(prealloc_cf); return err; out_put: ceph_mdsc_put_request(req); + ceph_free_cap_flush(prealloc_cf); return err; } diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c index 89e4305a94d4..8d73fe9d488b 100644 --- a/fs/ceph/mds_client.c +++ b/fs/ceph/mds_client.c @@ -1189,6 +1189,10 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap, } spin_unlock(&mdsc->cap_dirty_lock); + if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) { + list_add(&ci->i_prealloc_cap_flush->list, &to_remove); + ci->i_prealloc_cap_flush = NULL; + } } spin_unlock(&ci->i_ceph_lock); while (!list_empty(&to_remove)) { @@ -1196,7 +1200,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap, cf = list_first_entry(&to_remove, struct ceph_cap_flush, list); list_del(&cf->list); - kfree(cf); + ceph_free_cap_flush(cf); } while (drop--) iput(inode); diff --git a/fs/ceph/super.c b/fs/ceph/super.c index edeb83c43112..d1c833c321b9 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -622,6 +622,7 @@ static void destroy_fs_client(struct ceph_fs_client *fsc) */ struct kmem_cache *ceph_inode_cachep; struct kmem_cache *ceph_cap_cachep; +struct kmem_cache *ceph_cap_flush_cachep; struct kmem_cache *ceph_dentry_cachep; struct kmem_cache *ceph_file_cachep; @@ -647,6 +648,10 @@ static int __init init_caches(void) SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); if (ceph_cap_cachep == NULL) goto bad_cap; + ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush, + SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); + if (ceph_cap_flush_cachep == NULL) + goto bad_cap_flush; ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info, SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); @@ -665,6 +670,8 @@ static int __init init_caches(void) bad_file: kmem_cache_destroy(ceph_dentry_cachep); bad_dentry: + kmem_cache_destroy(ceph_cap_flush_cachep); +bad_cap_flush: kmem_cache_destroy(ceph_cap_cachep); bad_cap: kmem_cache_destroy(ceph_inode_cachep); @@ -681,6 +688,7 @@ static void destroy_caches(void) kmem_cache_destroy(ceph_inode_cachep); kmem_cache_destroy(ceph_cap_cachep); + kmem_cache_destroy(ceph_cap_flush_cachep); kmem_cache_destroy(ceph_dentry_cachep); kmem_cache_destroy(ceph_file_cachep); diff --git a/fs/ceph/super.h b/fs/ceph/super.h index e7f13f742357..4415e977d72b 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -309,6 +309,7 @@ struct ceph_inode_info { /* we need to track cap writeback on a per-cap-bit basis, to allow * overlapping, pipelined cap flushes to the mds. we can probably * reduce the tid to 8 bits if we're concerned about inode size. */ + struct ceph_cap_flush *i_prealloc_cap_flush; struct rb_root i_cap_flush_tree; wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ unsigned long i_hold_caps_min; /* jiffies */ @@ -578,7 +579,10 @@ static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) { return ci->i_dirty_caps | ci->i_flushing_caps; } -extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); +extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); +extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); +extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, + struct ceph_cap_flush **pcf); extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, struct ceph_cap *ocap, int mask); diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c index c6f7d9b82085..819163d8313b 100644 --- a/fs/ceph/xattr.c +++ b/fs/ceph/xattr.c @@ -912,6 +912,7 @@ int __ceph_setxattr(struct dentry *dentry, const char *name, struct ceph_vxattr *vxattr; struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc; + struct ceph_cap_flush *prealloc_cf = NULL; int issued; int err; int dirty = 0; @@ -950,6 +951,10 @@ int __ceph_setxattr(struct dentry *dentry, const char *name, if (!xattr) goto out; + prealloc_cf = ceph_alloc_cap_flush(); + if (!prealloc_cf) + goto out; + spin_lock(&ci->i_ceph_lock); retry: issued = __ceph_caps_issued(ci, NULL); @@ -991,7 +996,8 @@ retry: flags, value ? 1 : -1, &xattr); if (!err) { - dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL); + dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL, + &prealloc_cf); ci->i_xattrs.dirty = true; inode->i_ctime = CURRENT_TIME; } @@ -1001,6 +1007,7 @@ retry: up_read(&mdsc->snap_rwsem); if (dirty) __mark_inode_dirty(inode, dirty); + ceph_free_cap_flush(prealloc_cf); return err; do_sync: @@ -1010,6 +1017,7 @@ do_sync_unlocked: up_read(&mdsc->snap_rwsem); err = ceph_sync_setxattr(dentry, name, value, size, flags); out: + ceph_free_cap_flush(prealloc_cf); kfree(newname); kfree(newval); kfree(xattr); @@ -1062,6 +1070,7 @@ int __ceph_removexattr(struct dentry *dentry, const char *name) struct ceph_vxattr *vxattr; struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc; + struct ceph_cap_flush *prealloc_cf = NULL; int issued; int err; int required_blob_size; @@ -1079,6 +1088,10 @@ int __ceph_removexattr(struct dentry *dentry, const char *name) if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN)) goto do_sync_unlocked; + prealloc_cf = ceph_alloc_cap_flush(); + if (!prealloc_cf) + return -ENOMEM; + err = -ENOMEM; spin_lock(&ci->i_ceph_lock); retry: @@ -1120,7 +1133,8 @@ retry: err = __remove_xattr_by_name(ceph_inode(inode), name); - dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL); + dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL, + &prealloc_cf); ci->i_xattrs.dirty = true; inode->i_ctime = CURRENT_TIME; spin_unlock(&ci->i_ceph_lock); @@ -1128,12 +1142,14 @@ retry: up_read(&mdsc->snap_rwsem); if (dirty) __mark_inode_dirty(inode, dirty); + ceph_free_cap_flush(prealloc_cf); return err; do_sync: spin_unlock(&ci->i_ceph_lock); do_sync_unlocked: if (lock_snap_rwsem) up_read(&mdsc->snap_rwsem); + ceph_free_cap_flush(prealloc_cf); err = ceph_send_removexattr(dentry, name); return err; } diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h index d73a569f9bf5..9ebee53d3bf5 100644 --- a/include/linux/ceph/libceph.h +++ b/include/linux/ceph/libceph.h @@ -174,6 +174,7 @@ static inline int calc_pages_for(u64 off, u64 len) extern struct kmem_cache *ceph_inode_cachep; extern struct kmem_cache *ceph_cap_cachep; +extern struct kmem_cache *ceph_cap_flush_cachep; extern struct kmem_cache *ceph_dentry_cachep; extern struct kmem_cache *ceph_file_cachep; -- cgit v1.2.3 From b459be739f97e2062b2ba77cfe8ea198dbd58904 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Fri, 12 Jun 2015 13:21:07 +0300 Subject: crush: sync up with userspace .. up to ceph.git commit 1db1abc8328d ("crush: eliminate ad hoc diff between kernel and userspace"). This fixes a bunch of recently pulled coding style issues and makes includes a bit cleaner. A patch "crush:Make the function crush_ln static" from Nicholas Krause is folded in as crush_ln() has been made static in userspace as well. Signed-off-by: Ilya Dryomov --- include/linux/crush/crush.h | 40 +++++++++++- include/linux/crush/hash.h | 6 ++ include/linux/crush/mapper.h | 2 +- net/ceph/crush/crush.c | 13 ++-- net/ceph/crush/crush_ln_table.h | 32 +++++----- net/ceph/crush/hash.c | 8 ++- net/ceph/crush/mapper.c | 137 ++++++++++++++++++++++++++-------------- 7 files changed, 160 insertions(+), 78 deletions(-) (limited to 'include') diff --git a/include/linux/crush/crush.h b/include/linux/crush/crush.h index 48a1a7d100f1..48b49305716b 100644 --- a/include/linux/crush/crush.h +++ b/include/linux/crush/crush.h @@ -1,7 +1,11 @@ #ifndef CEPH_CRUSH_CRUSH_H #define CEPH_CRUSH_CRUSH_H -#include +#ifdef __KERNEL__ +# include +#else +# include "crush_compat.h" +#endif /* * CRUSH is a pseudo-random data distribution algorithm that @@ -20,7 +24,11 @@ #define CRUSH_MAGIC 0x00010000ul /* for detecting algorithm revisions */ #define CRUSH_MAX_DEPTH 10 /* max crush hierarchy depth */ +#define CRUSH_MAX_RULESET (1<<8) /* max crush ruleset number */ +#define CRUSH_MAX_RULES CRUSH_MAX_RULESET /* should be the same as max rulesets */ +#define CRUSH_MAX_DEVICE_WEIGHT (100u * 0x10000u) +#define CRUSH_MAX_BUCKET_WEIGHT (65535u * 0x10000u) #define CRUSH_ITEM_UNDEF 0x7ffffffe /* undefined result (internal use only) */ #define CRUSH_ITEM_NONE 0x7fffffff /* no result */ @@ -108,6 +116,15 @@ enum { }; extern const char *crush_bucket_alg_name(int alg); +/* + * although tree was a legacy algorithm, it has been buggy, so + * exclude it. + */ +#define CRUSH_LEGACY_ALLOWED_BUCKET_ALGS ( \ + (1 << CRUSH_BUCKET_UNIFORM) | \ + (1 << CRUSH_BUCKET_LIST) | \ + (1 << CRUSH_BUCKET_STRAW)) + struct crush_bucket { __s32 id; /* this'll be negative */ __u16 type; /* non-zero; type=0 is reserved for devices */ @@ -174,7 +191,7 @@ struct crush_map { /* choose local attempts using a fallback permutation before * re-descent */ __u32 choose_local_fallback_tries; - /* choose attempts before giving up */ + /* choose attempts before giving up */ __u32 choose_total_tries; /* attempt chooseleaf inner descent once for firstn mode; on * reject retry outer descent. Note that this does *not* @@ -187,6 +204,25 @@ struct crush_map { * that want to limit reshuffling, a value of 3 or 4 will make the * mappings line up a bit better with previous mappings. */ __u8 chooseleaf_vary_r; + +#ifndef __KERNEL__ + /* + * version 0 (original) of straw_calc has various flaws. version 1 + * fixes a few of them. + */ + __u8 straw_calc_version; + + /* + * allowed bucket algs is a bitmask, here the bit positions + * are CRUSH_BUCKET_*. note that these are *bits* and + * CRUSH_BUCKET_* values are not, so we need to or together (1 + * << CRUSH_BUCKET_WHATEVER). The 0th bit is not used to + * minimize confusion (bucket type values start at 1). + */ + __u32 allowed_bucket_algs; + + __u32 *choose_tries; +#endif }; diff --git a/include/linux/crush/hash.h b/include/linux/crush/hash.h index 91e884230d5d..d1d90258242e 100644 --- a/include/linux/crush/hash.h +++ b/include/linux/crush/hash.h @@ -1,6 +1,12 @@ #ifndef CEPH_CRUSH_HASH_H #define CEPH_CRUSH_HASH_H +#ifdef __KERNEL__ +# include +#else +# include "crush_compat.h" +#endif + #define CRUSH_HASH_RJENKINS1 0 #define CRUSH_HASH_DEFAULT CRUSH_HASH_RJENKINS1 diff --git a/include/linux/crush/mapper.h b/include/linux/crush/mapper.h index eab367446eea..5dfd5b1125d2 100644 --- a/include/linux/crush/mapper.h +++ b/include/linux/crush/mapper.h @@ -8,7 +8,7 @@ * LGPL2 */ -#include +#include "crush.h" extern int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size); extern int crush_do_rule(const struct crush_map *map, diff --git a/net/ceph/crush/crush.c b/net/ceph/crush/crush.c index 9d84ce4ea0df..80d7c3a97cb8 100644 --- a/net/ceph/crush/crush.c +++ b/net/ceph/crush/crush.c @@ -1,15 +1,11 @@ - #ifdef __KERNEL__ # include +# include #else -# include -# include -# define kfree(x) do { if (x) free(x); } while (0) -# define BUG_ON(x) assert(!(x)) +# include "crush_compat.h" +# include "crush.h" #endif -#include - const char *crush_bucket_alg_name(int alg) { switch (alg) { @@ -134,6 +130,9 @@ void crush_destroy(struct crush_map *map) kfree(map->rules); } +#ifndef __KERNEL__ + kfree(map->choose_tries); +#endif kfree(map); } diff --git a/net/ceph/crush/crush_ln_table.h b/net/ceph/crush/crush_ln_table.h index 6192c7fc958c..aae534c901a4 100644 --- a/net/ceph/crush/crush_ln_table.h +++ b/net/ceph/crush/crush_ln_table.h @@ -10,20 +10,20 @@ * */ -#if defined(__linux__) -#include -#elif defined(__FreeBSD__) -#include -#endif - #ifndef CEPH_CRUSH_LN_H #define CEPH_CRUSH_LN_H +#ifdef __KERNEL__ +# include +#else +# include "crush_compat.h" +#endif -// RH_LH_tbl[2*k] = 2^48/(1.0+k/128.0) -// RH_LH_tbl[2*k+1] = 2^48*log2(1.0+k/128.0) - -static int64_t __RH_LH_tbl[128*2+2] = { +/* + * RH_LH_tbl[2*k] = 2^48/(1.0+k/128.0) + * RH_LH_tbl[2*k+1] = 2^48*log2(1.0+k/128.0) + */ +static __s64 __RH_LH_tbl[128*2+2] = { 0x0001000000000000ll, 0x0000000000000000ll, 0x0000fe03f80fe040ll, 0x000002dfca16dde1ll, 0x0000fc0fc0fc0fc1ll, 0x000005b9e5a170b4ll, 0x0000fa232cf25214ll, 0x0000088e68ea899all, 0x0000f83e0f83e0f9ll, 0x00000b5d69bac77ell, 0x0000f6603d980f67ll, 0x00000e26fd5c8555ll, @@ -89,11 +89,12 @@ static int64_t __RH_LH_tbl[128*2+2] = { 0x0000820820820821ll, 0x0000fa2f045e7832ll, 0x000081848da8faf1ll, 0x0000fba577877d7dll, 0x0000810204081021ll, 0x0000fd1a708bbe11ll, 0x0000808080808081ll, 0x0000fe8df263f957ll, 0x0000800000000000ll, 0x0000ffff00000000ll, - }; - +}; - // LL_tbl[k] = 2^48*log2(1.0+k/2^15); -static int64_t __LL_tbl[256] = { +/* + * LL_tbl[k] = 2^48*log2(1.0+k/2^15) + */ +static __s64 __LL_tbl[256] = { 0x0000000000000000ull, 0x00000002e2a60a00ull, 0x000000070cb64ec5ull, 0x00000009ef50ce67ull, 0x0000000cd1e588fdull, 0x0000000fb4747e9cull, 0x0000001296fdaf5eull, 0x0000001579811b58ull, 0x000000185bfec2a1ull, 0x0000001b3e76a552ull, 0x0000001e20e8c380ull, 0x0000002103551d43ull, @@ -160,7 +161,4 @@ static int64_t __LL_tbl[256] = { 0x000002d4562d2ec6ull, 0x000002d73330209dull, 0x000002da102d63b0ull, 0x000002dced24f814ull, }; - - - #endif diff --git a/net/ceph/crush/hash.c b/net/ceph/crush/hash.c index 5bb63e37a8a1..ed123af49eba 100644 --- a/net/ceph/crush/hash.c +++ b/net/ceph/crush/hash.c @@ -1,6 +1,8 @@ - -#include -#include +#ifdef __KERNEL__ +# include +#else +# include "hash.h" +#endif /* * Robert Jenkins' function for mixing 32-bit values diff --git a/net/ceph/crush/mapper.c b/net/ceph/crush/mapper.c index 7568cb59b9e5..393bfb22d5bb 100644 --- a/net/ceph/crush/mapper.c +++ b/net/ceph/crush/mapper.c @@ -1,27 +1,31 @@ +/* + * Ceph - scalable distributed file system + * + * Copyright (C) 2015 Intel Corporation All Rights Reserved + * + * This is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License version 2.1, as published by the Free Software + * Foundation. See file COPYING. + * + */ #ifdef __KERNEL__ # include # include # include # include -# ifndef dprintk -# define dprintk(args...) -# endif +# include +# include #else -# include -# include -# include -# include -# define BUG_ON(x) assert(!(x)) -# define dprintk(args...) /* printf(args) */ -# define kmalloc(x, f) malloc(x) -# define kfree(x) free(x) +# include "crush_compat.h" +# include "crush.h" +# include "hash.h" #endif - -#include -#include #include "crush_ln_table.h" +#define dprintk(args...) /* printf(args) */ + /* * Implement the core CRUSH mapping algorithm. */ @@ -139,7 +143,7 @@ static int bucket_list_choose(struct crush_bucket_list *bucket, int i; for (i = bucket->h.size-1; i >= 0; i--) { - __u64 w = crush_hash32_4(bucket->h.hash,x, bucket->h.items[i], + __u64 w = crush_hash32_4(bucket->h.hash, x, bucket->h.items[i], r, bucket->h.id); w &= 0xffff; dprintk("list_choose i=%d x=%d r=%d item %d weight %x " @@ -238,43 +242,46 @@ static int bucket_straw_choose(struct crush_bucket_straw *bucket, return bucket->h.items[high]; } -// compute 2^44*log2(input+1) -uint64_t crush_ln(unsigned xin) +/* compute 2^44*log2(input+1) */ +static __u64 crush_ln(unsigned int xin) { - unsigned x=xin, x1; - int iexpon, index1, index2; - uint64_t RH, LH, LL, xl64, result; + unsigned int x = xin, x1; + int iexpon, index1, index2; + __u64 RH, LH, LL, xl64, result; - x++; + x++; - // normalize input - iexpon = 15; - while(!(x&0x18000)) { x<<=1; iexpon--; } + /* normalize input */ + iexpon = 15; + while (!(x & 0x18000)) { + x <<= 1; + iexpon--; + } - index1 = (x>>8)<<1; - // RH ~ 2^56/index1 - RH = __RH_LH_tbl[index1 - 256]; - // LH ~ 2^48 * log2(index1/256) - LH = __RH_LH_tbl[index1 + 1 - 256]; + index1 = (x >> 8) << 1; + /* RH ~ 2^56/index1 */ + RH = __RH_LH_tbl[index1 - 256]; + /* LH ~ 2^48 * log2(index1/256) */ + LH = __RH_LH_tbl[index1 + 1 - 256]; - // RH*x ~ 2^48 * (2^15 + xf), xf<2^8 - xl64 = (int64_t)x * RH; - xl64 >>= 48; - x1 = xl64; + /* RH*x ~ 2^48 * (2^15 + xf), xf<2^8 */ + xl64 = (__s64)x * RH; + xl64 >>= 48; + x1 = xl64; - result = iexpon; - result <<= (12 + 32); + result = iexpon; + result <<= (12 + 32); - index2 = x1 & 0xff; - // LL ~ 2^48*log2(1.0+index2/2^15) - LL = __LL_tbl[index2]; + index2 = x1 & 0xff; + /* LL ~ 2^48*log2(1.0+index2/2^15) */ + LL = __LL_tbl[index2]; - LH = LH + LL; + LH = LH + LL; - LH >>= (48-12 - 32); - result += LH; + LH >>= (48 - 12 - 32); + result += LH; - return result; + return result; } @@ -290,9 +297,9 @@ uint64_t crush_ln(unsigned xin) static int bucket_straw2_choose(struct crush_bucket_straw2 *bucket, int x, int r) { - unsigned i, high = 0; - unsigned u; - unsigned w; + unsigned int i, high = 0; + unsigned int u; + unsigned int w; __s64 ln, draw, high_draw = 0; for (i = 0; i < bucket->h.size; i++) { @@ -567,6 +574,10 @@ reject: out[outpos] = item; outpos++; count--; +#ifndef __KERNEL__ + if (map->choose_tries && ftotal <= map->choose_total_tries) + map->choose_tries[ftotal]++; +#endif } dprintk("CHOOSE returns %d\n", outpos); @@ -610,6 +621,20 @@ static void crush_choose_indep(const struct crush_map *map, } for (ftotal = 0; left > 0 && ftotal < tries; ftotal++) { +#ifdef DEBUG_INDEP + if (out2 && ftotal) { + dprintk("%u %d a: ", ftotal, left); + for (rep = outpos; rep < endpos; rep++) { + dprintk(" %d", out[rep]); + } + dprintk("\n"); + dprintk("%u %d b: ", ftotal, left); + for (rep = outpos; rep < endpos; rep++) { + dprintk(" %d", out2[rep]); + } + dprintk("\n"); + } +#endif for (rep = outpos; rep < endpos; rep++) { if (out[rep] != CRUSH_ITEM_UNDEF) continue; @@ -726,6 +751,24 @@ static void crush_choose_indep(const struct crush_map *map, out2[rep] = CRUSH_ITEM_NONE; } } +#ifndef __KERNEL__ + if (map->choose_tries && ftotal <= map->choose_total_tries) + map->choose_tries[ftotal]++; +#endif +#ifdef DEBUG_INDEP + if (out2) { + dprintk("%u %d a: ", ftotal, left); + for (rep = outpos; rep < endpos; rep++) { + dprintk(" %d", out[rep]); + } + dprintk("\n"); + dprintk("%u %d b: ", ftotal, left); + for (rep = outpos; rep < endpos; rep++) { + dprintk(" %d", out2[rep]); + } + dprintk("\n"); + } +#endif } /** @@ -884,7 +927,7 @@ int crush_do_rule(const struct crush_map *map, 0); } else { out_size = ((numrep < (result_max-osize)) ? - numrep : (result_max-osize)); + numrep : (result_max-osize)); crush_choose_indep( map, map->buckets[-1-w[i]], @@ -930,5 +973,3 @@ int crush_do_rule(const struct crush_map *map, } return result_len; } - - -- cgit v1.2.3