From 3d14c5d2b6e15c21d8e5467dc62d33127c23a644 Mon Sep 17 00:00:00 2001 From: Yehuda Sadeh Date: Tue, 6 Apr 2010 15:14:15 -0700 Subject: ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil --- net/ceph/pagelist.c | 65 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 65 insertions(+) create mode 100644 net/ceph/pagelist.c (limited to 'net/ceph/pagelist.c') diff --git a/net/ceph/pagelist.c b/net/ceph/pagelist.c new file mode 100644 index 00000000000..3b146cfac18 --- /dev/null +++ b/net/ceph/pagelist.c @@ -0,0 +1,65 @@ + +#include +#include +#include +#include +#include + +static void ceph_pagelist_unmap_tail(struct ceph_pagelist *pl) +{ + struct page *page = list_entry(pl->head.prev, struct page, + lru); + kunmap(page); +} + +int ceph_pagelist_release(struct ceph_pagelist *pl) +{ + if (pl->mapped_tail) + ceph_pagelist_unmap_tail(pl); + + while (!list_empty(&pl->head)) { + struct page *page = list_first_entry(&pl->head, struct page, + lru); + list_del(&page->lru); + __free_page(page); + } + return 0; +} +EXPORT_SYMBOL(ceph_pagelist_release); + +static int ceph_pagelist_addpage(struct ceph_pagelist *pl) +{ + struct page *page = __page_cache_alloc(GFP_NOFS); + if (!page) + return -ENOMEM; + pl->room += PAGE_SIZE; + list_add_tail(&page->lru, &pl->head); + if (pl->mapped_tail) + ceph_pagelist_unmap_tail(pl); + pl->mapped_tail = kmap(page); + return 0; +} + +int ceph_pagelist_append(struct ceph_pagelist *pl, const void *buf, size_t len) +{ + while (pl->room < len) { + size_t bit = pl->room; + int ret; + + memcpy(pl->mapped_tail + (pl->length & ~PAGE_CACHE_MASK), + buf, bit); + pl->length += bit; + pl->room -= bit; + buf += bit; + len -= bit; + ret = ceph_pagelist_addpage(pl); + if (ret) + return ret; + } + + memcpy(pl->mapped_tail + (pl->length & ~PAGE_CACHE_MASK), buf, len); + pl->length += len; + pl->room -= len; + return 0; +} +EXPORT_SYMBOL(ceph_pagelist_append); -- cgit v1.2.3-70-g09d2 From ac0b74d8a1ced8ea86147467daf06b15b130dd94 Mon Sep 17 00:00:00 2001 From: Greg Farnum Date: Fri, 17 Sep 2010 10:10:55 -0700 Subject: ceph: add pagelist_reserve, pagelist_truncate, pagelist_set_cursor These facilitate preallocation of pages so that we can encode into the pagelist in an atomic context. Signed-off-by: Greg Farnum Signed-off-by: Sage Weil --- include/linux/ceph/pagelist.h | 21 +++++++++ net/ceph/pagelist.c | 106 ++++++++++++++++++++++++++++++++++++++---- 2 files changed, 118 insertions(+), 9 deletions(-) (limited to 'net/ceph/pagelist.c') diff --git a/include/linux/ceph/pagelist.h b/include/linux/ceph/pagelist.h index cc9327aa1c9..9660d6b0a35 100644 --- a/include/linux/ceph/pagelist.h +++ b/include/linux/ceph/pagelist.h @@ -8,6 +8,14 @@ struct ceph_pagelist { void *mapped_tail; size_t length; size_t room; + struct list_head free_list; + size_t num_pages_free; +}; + +struct ceph_pagelist_cursor { + struct ceph_pagelist *pl; /* pagelist, for error checking */ + struct list_head *page_lru; /* page in list */ + size_t room; /* room remaining to reset to */ }; static inline void ceph_pagelist_init(struct ceph_pagelist *pl) @@ -16,11 +24,24 @@ static inline void ceph_pagelist_init(struct ceph_pagelist *pl) pl->mapped_tail = NULL; pl->length = 0; pl->room = 0; + INIT_LIST_HEAD(&pl->free_list); + pl->num_pages_free = 0; } + extern int ceph_pagelist_release(struct ceph_pagelist *pl); extern int ceph_pagelist_append(struct ceph_pagelist *pl, const void *d, size_t l); +extern int ceph_pagelist_reserve(struct ceph_pagelist *pl, size_t space); + +extern int ceph_pagelist_free_reserve(struct ceph_pagelist *pl); + +extern void ceph_pagelist_set_cursor(struct ceph_pagelist *pl, + struct ceph_pagelist_cursor *c); + +extern int ceph_pagelist_truncate(struct ceph_pagelist *pl, + struct ceph_pagelist_cursor *c); + static inline int ceph_pagelist_encode_64(struct ceph_pagelist *pl, u64 v) { __le64 ev = cpu_to_le64(v); diff --git a/net/ceph/pagelist.c b/net/ceph/pagelist.c index 3b146cfac18..b8cbc456d0b 100644 --- a/net/ceph/pagelist.c +++ b/net/ceph/pagelist.c @@ -7,35 +7,42 @@ static void ceph_pagelist_unmap_tail(struct ceph_pagelist *pl) { - struct page *page = list_entry(pl->head.prev, struct page, - lru); - kunmap(page); + if (pl->mapped_tail) { + struct page *page = list_entry(pl->head.prev, struct page, lru); + kunmap(page); + pl->mapped_tail = NULL; + } } int ceph_pagelist_release(struct ceph_pagelist *pl) { - if (pl->mapped_tail) - ceph_pagelist_unmap_tail(pl); - + ceph_pagelist_unmap_tail(pl); while (!list_empty(&pl->head)) { struct page *page = list_first_entry(&pl->head, struct page, lru); list_del(&page->lru); __free_page(page); } + ceph_pagelist_free_reserve(pl); return 0; } EXPORT_SYMBOL(ceph_pagelist_release); static int ceph_pagelist_addpage(struct ceph_pagelist *pl) { - struct page *page = __page_cache_alloc(GFP_NOFS); + struct page *page; + + if (!pl->num_pages_free) { + page = __page_cache_alloc(GFP_NOFS); + } else { + page = list_first_entry(&pl->free_list, struct page, lru); + list_del(&page->lru); + } if (!page) return -ENOMEM; pl->room += PAGE_SIZE; + ceph_pagelist_unmap_tail(pl); list_add_tail(&page->lru, &pl->head); - if (pl->mapped_tail) - ceph_pagelist_unmap_tail(pl); pl->mapped_tail = kmap(page); return 0; } @@ -63,3 +70,84 @@ int ceph_pagelist_append(struct ceph_pagelist *pl, const void *buf, size_t len) return 0; } EXPORT_SYMBOL(ceph_pagelist_append); + +/** + * Allocate enough pages for a pagelist to append the given amount + * of data without without allocating. + * Returns: 0 on success, -ENOMEM on error. + */ +int ceph_pagelist_reserve(struct ceph_pagelist *pl, size_t space) +{ + if (space <= pl->room) + return 0; + space -= pl->room; + space = (space + PAGE_SIZE - 1) >> PAGE_SHIFT; /* conv to num pages */ + + while (space > pl->num_pages_free) { + struct page *page = __page_cache_alloc(GFP_NOFS); + if (!page) + return -ENOMEM; + list_add_tail(&page->lru, &pl->free_list); + ++pl->num_pages_free; + } + return 0; +} +EXPORT_SYMBOL(ceph_pagelist_reserve); + +/** + * Free any pages that have been preallocated. + */ +int ceph_pagelist_free_reserve(struct ceph_pagelist *pl) +{ + while (!list_empty(&pl->free_list)) { + struct page *page = list_first_entry(&pl->free_list, + struct page, lru); + list_del(&page->lru); + __free_page(page); + --pl->num_pages_free; + } + BUG_ON(pl->num_pages_free); + return 0; +} +EXPORT_SYMBOL(ceph_pagelist_free_reserve); + +/** + * Create a truncation point. + */ +void ceph_pagelist_set_cursor(struct ceph_pagelist *pl, + struct ceph_pagelist_cursor *c) +{ + c->pl = pl; + c->page_lru = pl->head.prev; + c->room = pl->room; +} +EXPORT_SYMBOL(ceph_pagelist_set_cursor); + +/** + * Truncate a pagelist to the given point. Move extra pages to reserve. + * This won't sleep. + * Returns: 0 on success, + * -EINVAL if the pagelist doesn't match the trunc point pagelist + */ +int ceph_pagelist_truncate(struct ceph_pagelist *pl, + struct ceph_pagelist_cursor *c) +{ + struct page *page; + + if (pl != c->pl) + return -EINVAL; + ceph_pagelist_unmap_tail(pl); + while (pl->head.prev != c->page_lru) { + page = list_entry(pl->head.prev, struct page, lru); + list_del(&page->lru); /* remove from pagelist */ + list_add_tail(&page->lru, &pl->free_list); /* add to reserve */ + ++pl->num_pages_free; + } + pl->room = c->room; + if (!list_empty(&pl->head)) { + page = list_entry(pl->head.prev, struct page, lru); + pl->mapped_tail = kmap(page); + } + return 0; +} +EXPORT_SYMBOL(ceph_pagelist_truncate); -- cgit v1.2.3-70-g09d2 From 240634e9b3a8ae2a9e2b7effeea3b9a5b30d68fb Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Tue, 5 Oct 2010 12:03:23 -0700 Subject: ceph: fix num_pages_free accounting in pagelist Decrement the free page counter when removing a page from the free_list. Signed-off-by: Sage Weil --- net/ceph/pagelist.c | 1 + 1 file changed, 1 insertion(+) (limited to 'net/ceph/pagelist.c') diff --git a/net/ceph/pagelist.c b/net/ceph/pagelist.c index b8cbc456d0b..13cb409a7bb 100644 --- a/net/ceph/pagelist.c +++ b/net/ceph/pagelist.c @@ -37,6 +37,7 @@ static int ceph_pagelist_addpage(struct ceph_pagelist *pl) } else { page = list_first_entry(&pl->free_list, struct page, lru); list_del(&page->lru); + --pl->num_pages_free; } if (!page) return -ENOMEM; -- cgit v1.2.3-70-g09d2