adulau/aha
1
0
Fork 0
mirror of https://github.com/adulau/aha.git synced 2024-12-27 19:26:25 +00:00
Code Issues Projects Releases Packages Wiki Activity

exofs: Move all operations to an io_engine

Browse source 
In anticipation for multi-device operations, we separate osd operations
into an abstract I/O API. Currently only one device is used but later
when adding more devices, we will drive all devices in parallel according
to a "data_map" that describes how data is arranged on multiple devices.
The file system level operates, like before, as if there is one object
(inode-number) and an i_size. The io engine will split this to the same
object-number but on multiple device.

At first we introduce Mirror (raid 1) layout. But at the final outcome
we intend to fully implement the pNFS-Objects data-map, including
raid 0,4,5,6 over mirrored devices, over multiple device-groups. And
more. See: http://tools.ietf.org/html/draft-ietf-nfsv4-pnfs-obj-12

* Define an io_state based API for accessing osd storage devices
  in an abstract way.
  Usage:
	First a caller allocates an io state with:
		exofs_get_io_state(struct exofs_sb_info *sbi,
				   struct exofs_io_state** ios);

	Then calles one of:
		exofs_sbi_create(struct exofs_io_state *ios);
		exofs_sbi_remove(struct exofs_io_state *ios);
		exofs_sbi_write(struct exofs_io_state *ios);
		exofs_sbi_read(struct exofs_io_state *ios);
		exofs_oi_truncate(struct exofs_i_info *oi, u64 new_len);

	And when done
		exofs_put_io_state(struct exofs_io_state *ios);

* Convert all source files to use this new API
* Convert from bio_alloc to bio_kmalloc
* In io engine we make use of the now fixed osd_req_decode_sense

There are no functional changes or on disk additions after this patch.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
This commit is contained in:
Boaz Harrosh 2009-11-08 14:54:08 +02:00
parent 8ce9bdd1fb
commit 06886a5a3d
5 changed files with 646 additions and 352 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
fs/exofs/common.h
View file

@ -155,22 +155,4 @@ enum {
(((name_len) + offsetof(struct exofs_dir_entry, name) + \ (((name_len) + offsetof(struct exofs_dir_entry, name) + \
EXOFS_DIR_ROUND) & ~EXOFS_DIR_ROUND) EXOFS_DIR_ROUND) & ~EXOFS_DIR_ROUND)
/*************************
* function declarations *
*************************/
/* osd.c */
void exofs_make_credential(u8 cred_a[OSD_CAP_LEN],
const struct osd_obj_id *obj);
int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid);
static inline int exofs_check_ok(struct osd_request *or)
{
return exofs_check_ok_resid(or, NULL, NULL);
}
int exofs_sync_op(struct osd_request *or, int timeout, u8 *cred);
int exofs_async_op(struct osd_request *or,
osd_req_done_fn *async_done, void *caller_context, u8 *cred);
int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr);
#endif /*ifndef __EXOFS_COM_H__*/ #endif /*ifndef __EXOFS_COM_H__*/

87
fs/exofs/exofs.h
View file

@ -30,14 +30,13 @@
* along with exofs; if not, write to the Free Software * along with exofs; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/ */
#ifndef __EXOFS_H__
#define __EXOFS_H__
#include <linux/fs.h> #include <linux/fs.h>
#include <linux/time.h> #include <linux/time.h>
#include "common.h" #include "common.h"
#ifndef __EXOFS_H__
#define __EXOFS_H__
#define EXOFS_ERR(fmt, a...) printk(KERN_ERR "exofs: " fmt, ##a) #define EXOFS_ERR(fmt, a...) printk(KERN_ERR "exofs: " fmt, ##a)
#ifdef CONFIG_EXOFS_DEBUG #ifdef CONFIG_EXOFS_DEBUG
@ -56,6 +55,7 @@
*/ */
struct exofs_sb_info { struct exofs_sb_info {
struct osd_dev *s_dev; /* returned by get_osd_dev */ struct osd_dev *s_dev; /* returned by get_osd_dev */
struct exofs_fscb s_fscb; /* Written often, pre-allocate*/
osd_id s_pid; /* partition ID of file system*/ osd_id s_pid; /* partition ID of file system*/
int s_timeout; /* timeout for OSD operations */ int s_timeout; /* timeout for OSD operations */
uint64_t s_nextid; /* highest object ID used */ uint64_t s_nextid; /* highest object ID used */
@ -79,6 +79,50 @@ struct exofs_i_info {
struct inode vfs_inode; /* normal in-memory inode */ struct inode vfs_inode; /* normal in-memory inode */
}; };
static inline osd_id exofs_oi_objno(struct exofs_i_info *oi)
{
return oi->vfs_inode.i_ino + EXOFS_OBJ_OFF;
}
struct exofs_io_state;
typedef void (*exofs_io_done_fn)(struct exofs_io_state *or, void *private);
struct exofs_io_state {
struct kref kref;
void *private;
exofs_io_done_fn done;
struct exofs_sb_info *sbi;
struct osd_obj_id obj;
u8 *cred;
/* Global read/write IO*/
loff_t offset;
unsigned long length;
void *kern_buff;
struct bio *bio;
/* Attributes */
unsigned in_attr_len;
struct osd_attr *in_attr;
unsigned out_attr_len;
struct osd_attr *out_attr;
/* Variable array of size numdevs */
unsigned numdevs;
struct exofs_per_dev_state {
struct osd_request *or;
struct bio *bio;
} per_dev[];
};
static inline unsigned exofs_io_state_size(unsigned numdevs)
{
return sizeof(struct exofs_io_state) +
sizeof(struct exofs_per_dev_state) * numdevs;
}
/* /*
* our inode flags * our inode flags
*/ */
@ -130,6 +174,42 @@ static inline struct exofs_i_info *exofs_i(struct inode *inode)
/************************* /*************************
* function declarations * * function declarations *
*************************/ *************************/
/* ios.c */
void exofs_make_credential(u8 cred_a[OSD_CAP_LEN],
const struct osd_obj_id *obj);
int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
u64 offset, void *p, unsigned length);
int exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** ios);
void exofs_put_io_state(struct exofs_io_state *ios);
int exofs_check_io(struct exofs_io_state *ios, u64 *resid);
int exofs_sbi_create(struct exofs_io_state *ios);
int exofs_sbi_remove(struct exofs_io_state *ios);
int exofs_sbi_write(struct exofs_io_state *ios);
int exofs_sbi_read(struct exofs_io_state *ios);
int extract_attr_from_ios(struct exofs_io_state *ios, struct osd_attr *attr);
int exofs_oi_truncate(struct exofs_i_info *oi, u64 new_len);
static inline int exofs_oi_write(struct exofs_i_info *oi,
struct exofs_io_state *ios)
{
ios->obj.id = exofs_oi_objno(oi);
ios->cred = oi->i_cred;
return exofs_sbi_write(ios);
}
static inline int exofs_oi_read(struct exofs_i_info *oi,
struct exofs_io_state *ios)
{
ios->obj.id = exofs_oi_objno(oi);
ios->cred = oi->i_cred;
return exofs_sbi_read(ios);
}
/* inode.c */ /* inode.c */
void exofs_truncate(struct inode *inode); void exofs_truncate(struct inode *inode);
int exofs_setattr(struct dentry *, struct iattr *); int exofs_setattr(struct dentry *, struct iattr *);
@ -169,6 +249,7 @@ extern const struct file_operations exofs_file_operations;
/* inode.c */ /* inode.c */
extern const struct address_space_operations exofs_aops; extern const struct address_space_operations exofs_aops;
extern const struct osd_attr g_attr_logical_length;
/* namei.c */ /* namei.c */
extern const struct inode_operations exofs_dir_inode_operations; extern const struct inode_operations exofs_dir_inode_operations;

383
fs/exofs/inode.c
View file

@ -37,17 +37,18 @@
#include "exofs.h" #include "exofs.h"
#ifdef CONFIG_EXOFS_DEBUG
# define EXOFS_DEBUG_OBJ_ISIZE 1
#endif
#define EXOFS_DBGMSG2(M...) do {} while (0) #define EXOFS_DBGMSG2(M...) do {} while (0)
enum { BIO_MAX_PAGES_KMALLOC =
(PAGE_SIZE - sizeof(struct bio)) / sizeof(struct bio_vec),
};
struct page_collect { struct page_collect {
struct exofs_sb_info *sbi; struct exofs_sb_info *sbi;
struct request_queue *req_q; struct request_queue *req_q;
struct inode *inode; struct inode *inode;
unsigned expected_pages; unsigned expected_pages;
struct exofs_io_state *ios;
struct bio *bio; struct bio *bio;
unsigned nr_pages; unsigned nr_pages;
@ -56,7 +57,7 @@ struct page_collect {
}; };
static void _pcol_init(struct page_collect *pcol, unsigned expected_pages, static void _pcol_init(struct page_collect *pcol, unsigned expected_pages,
struct inode *inode) struct inode *inode)
{ {
struct exofs_sb_info *sbi = inode->i_sb->s_fs_info; struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
@ -65,13 +66,11 @@ static void _pcol_init(struct page_collect *pcol, unsigned expected_pages,
pcol->inode = inode; pcol->inode = inode;
pcol->expected_pages = expected_pages; pcol->expected_pages = expected_pages;
pcol->ios = NULL;
pcol->bio = NULL; pcol->bio = NULL;
pcol->nr_pages = 0; pcol->nr_pages = 0;
pcol->length = 0; pcol->length = 0;
pcol->pg_first = -1; pcol->pg_first = -1;
EXOFS_DBGMSG("_pcol_init ino=0x%lx expected_pages=%u\n", inode->i_ino,
expected_pages);
} }
static void _pcol_reset(struct page_collect *pcol) static void _pcol_reset(struct page_collect *pcol)
@ -82,35 +81,49 @@ static void _pcol_reset(struct page_collect *pcol)
pcol->nr_pages = 0; pcol->nr_pages = 0;
pcol->length = 0; pcol->length = 0;
pcol->pg_first = -1; pcol->pg_first = -1;
EXOFS_DBGMSG("_pcol_reset ino=0x%lx expected_pages=%u\n", pcol->ios = NULL;
pcol->inode->i_ino, pcol->expected_pages);
/* this is probably the end of the loop but in writes /* this is probably the end of the loop but in writes
* it might not end here. don't be left with nothing * it might not end here. don't be left with nothing
*/ */
if (!pcol->expected_pages) if (!pcol->expected_pages)
pcol->expected_pages = 128; pcol->expected_pages = BIO_MAX_PAGES_KMALLOC;
} }
static int pcol_try_alloc(struct page_collect *pcol) static int pcol_try_alloc(struct page_collect *pcol)
{ {
int pages = min_t(unsigned, pcol->expected_pages, BIO_MAX_PAGES); int pages = min_t(unsigned, pcol->expected_pages,
BIO_MAX_PAGES_KMALLOC);
if (!pcol->ios) { /* First time allocate io_state */
int ret = exofs_get_io_state(pcol->sbi, &pcol->ios);
if (ret)
return ret;
}
for (; pages; pages >>= 1) { for (; pages; pages >>= 1) {
pcol->bio = bio_alloc(GFP_KERNEL, pages); pcol->bio = bio_kmalloc(GFP_KERNEL, pages);
if (likely(pcol->bio)) if (likely(pcol->bio))
return 0; return 0;
} }
EXOFS_ERR("Failed to kcalloc expected_pages=%u\n", EXOFS_ERR("Failed to bio_kmalloc expected_pages=%u\n",
pcol->expected_pages); pcol->expected_pages);
return -ENOMEM; return -ENOMEM;
} }
static void pcol_free(struct page_collect *pcol) static void pcol_free(struct page_collect *pcol)
{ {
bio_put(pcol->bio); if (pcol->bio) {
pcol->bio = NULL; bio_put(pcol->bio);
pcol->bio = NULL;
}
if (pcol->ios) {
exofs_put_io_state(pcol->ios);
pcol->ios = NULL;
}
} }
static int pcol_add_page(struct page_collect *pcol, struct page *page, static int pcol_add_page(struct page_collect *pcol, struct page *page,
@ -163,22 +176,17 @@ static void update_write_page(struct page *page, int ret)
/* Called at the end of reads, to optionally unlock pages and update their /* Called at the end of reads, to optionally unlock pages and update their
* status. * status.
*/ */
static int __readpages_done(struct osd_request *or, struct page_collect *pcol, static int __readpages_done(struct page_collect *pcol, bool do_unlock)
bool do_unlock)
{ {
struct bio_vec *bvec; struct bio_vec *bvec;
int i; int i;
u64 resid; u64 resid;
u64 good_bytes; u64 good_bytes;
u64 length = 0; u64 length = 0;
int ret = exofs_check_ok_resid(or, &resid, NULL); int ret = exofs_check_io(pcol->ios, &resid);
osd_end_request(or);
if (likely(!ret)) if (likely(!ret))
good_bytes = pcol->length; good_bytes = pcol->length;
else if (!resid)
good_bytes = 0;
else else
good_bytes = pcol->length - resid; good_bytes = pcol->length - resid;
@ -216,13 +224,13 @@ static int __readpages_done(struct osd_request *or, struct page_collect *pcol,
} }
/* callback of async reads */ /* callback of async reads */
static void readpages_done(struct osd_request *or, void *p) static void readpages_done(struct exofs_io_state *ios, void *p)
{ {
struct page_collect *pcol = p; struct page_collect *pcol = p;
__readpages_done(or, pcol, true); __readpages_done(pcol, true);
atomic_dec(&pcol->sbi->s_curr_pending); atomic_dec(&pcol->sbi->s_curr_pending);
kfree(p); kfree(pcol);
} }
static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw) static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw)
@ -240,17 +248,13 @@ static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw)
unlock_page(page); unlock_page(page);
} }
pcol_free(pcol);
} }
static int read_exec(struct page_collect *pcol, bool is_sync) static int read_exec(struct page_collect *pcol, bool is_sync)
{ {
struct exofs_i_info *oi = exofs_i(pcol->inode); struct exofs_i_info *oi = exofs_i(pcol->inode);
struct osd_obj_id obj = {pcol->sbi->s_pid, struct exofs_io_state *ios = pcol->ios;
pcol->inode->i_ino + EXOFS_OBJ_OFF};
struct osd_request *or = NULL;
struct page_collect *pcol_copy = NULL; struct page_collect *pcol_copy = NULL;
loff_t i_start = pcol->pg_first << PAGE_CACHE_SHIFT;
int ret; int ret;
if (!pcol->bio) if (!pcol->bio)
@ -259,17 +263,13 @@ static int read_exec(struct page_collect *pcol, bool is_sync)
/* see comment in _readpage() about sync reads */ /* see comment in _readpage() about sync reads */
WARN_ON(is_sync && (pcol->nr_pages != 1)); WARN_ON(is_sync && (pcol->nr_pages != 1));
or = osd_start_request(pcol->sbi->s_dev, GFP_KERNEL); ios->bio = pcol->bio;
if (unlikely(!or)) { ios->length = pcol->length;
ret = -ENOMEM; ios->offset = pcol->pg_first << PAGE_CACHE_SHIFT;
goto err;
}
osd_req_read(or, &obj, i_start, pcol->bio, pcol->length);
if (is_sync) { if (is_sync) {
exofs_sync_op(or, pcol->sbi->s_timeout, oi->i_cred); exofs_oi_read(oi, pcol->ios);
return __readpages_done(or, pcol, false); return __readpages_done(pcol, false);
} }
pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL); pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL);
@ -279,14 +279,16 @@ static int read_exec(struct page_collect *pcol, bool is_sync)
} }
*pcol_copy = *pcol; *pcol_copy = *pcol;
ret = exofs_async_op(or, readpages_done, pcol_copy, oi->i_cred); ios->done = readpages_done;
ios->private = pcol_copy;
ret = exofs_oi_read(oi, ios);
if (unlikely(ret)) if (unlikely(ret))
goto err; goto err;
atomic_inc(&pcol->sbi->s_curr_pending); atomic_inc(&pcol->sbi->s_curr_pending);
EXOFS_DBGMSG("read_exec obj=0x%llx start=0x%llx length=0x%lx\n", EXOFS_DBGMSG("read_exec obj=0x%llx start=0x%llx length=0x%lx\n",
obj.id, _LLU(i_start), pcol->length); ios->obj.id, _LLU(ios->offset), pcol->length);
/* pages ownership was passed to pcol_copy */ /* pages ownership was passed to pcol_copy */
_pcol_reset(pcol); _pcol_reset(pcol);
@ -295,12 +297,10 @@ static int read_exec(struct page_collect *pcol, bool is_sync)
err: err:
if (!is_sync) if (!is_sync)
_unlock_pcol_pages(pcol, ret, READ); _unlock_pcol_pages(pcol, ret, READ);
else /* Pages unlocked by caller in sync mode only free bio */
pcol_free(pcol); pcol_free(pcol);
kfree(pcol_copy); kfree(pcol_copy);
if (or)
osd_end_request(or);
return ret; return ret;
} }
@ -421,9 +421,8 @@ static int _readpage(struct page *page, bool is_sync)
_pcol_init(&pcol, 1, page->mapping->host); _pcol_init(&pcol, 1, page->mapping->host);
/* readpage_strip might call read_exec(,async) inside at several places /* readpage_strip might call read_exec(,is_sync==false) at several
* but this is safe for is_async=0 since read_exec will not do anything * places but not if we have a single page.
* when we have a single page.
*/ */
ret = readpage_strip(&pcol, page); ret = readpage_strip(&pcol, page);
if (ret) { if (ret) {
@ -442,8 +441,8 @@ static int exofs_readpage(struct file *file, struct page *page)
return _readpage(page, false); return _readpage(page, false);
} }
/* Callback for osd_write. All writes are asynchronouse */ /* Callback for osd_write. All writes are asynchronous */
static void writepages_done(struct osd_request *or, void *p) static void writepages_done(struct exofs_io_state *ios, void *p)
{ {
struct page_collect *pcol = p; struct page_collect *pcol = p;
struct bio_vec *bvec; struct bio_vec *bvec;
@ -451,16 +450,12 @@ static void writepages_done(struct osd_request *or, void *p)
u64 resid; u64 resid;
u64 good_bytes; u64 good_bytes;
u64 length = 0; u64 length = 0;
int ret = exofs_check_io(ios, &resid);
int ret = exofs_check_ok_resid(or, NULL, &resid);
osd_end_request(or);
atomic_dec(&pcol->sbi->s_curr_pending); atomic_dec(&pcol->sbi->s_curr_pending);
if (likely(!ret)) if (likely(!ret))
good_bytes = pcol->length; good_bytes = pcol->length;
else if (!resid)
good_bytes = 0;
else else
good_bytes = pcol->length - resid; good_bytes = pcol->length - resid;
@ -498,23 +493,13 @@ static void writepages_done(struct osd_request *or, void *p)
static int write_exec(struct page_collect *pcol) static int write_exec(struct page_collect *pcol)
{ {
struct exofs_i_info *oi = exofs_i(pcol->inode); struct exofs_i_info *oi = exofs_i(pcol->inode);
struct osd_obj_id obj = {pcol->sbi->s_pid, struct exofs_io_state *ios = pcol->ios;
pcol->inode->i_ino + EXOFS_OBJ_OFF};
struct osd_request *or = NULL;
struct page_collect *pcol_copy = NULL; struct page_collect *pcol_copy = NULL;
loff_t i_start = pcol->pg_first << PAGE_CACHE_SHIFT;
int ret; int ret;
if (!pcol->bio) if (!pcol->bio)
return 0; return 0;
or = osd_start_request(pcol->sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("write_exec: Faild to osd_start_request()\n");
ret = -ENOMEM;
goto err;
}
pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL); pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL);
if (!pcol_copy) { if (!pcol_copy) {
EXOFS_ERR("write_exec: Faild to kmalloc(pcol)\n"); EXOFS_ERR("write_exec: Faild to kmalloc(pcol)\n");
@ -525,16 +510,22 @@ static int write_exec(struct page_collect *pcol)
*pcol_copy = *pcol; *pcol_copy = *pcol;
pcol_copy->bio->bi_rw |= (1 << BIO_RW); /* FIXME: bio_set_dir() */ pcol_copy->bio->bi_rw |= (1 << BIO_RW); /* FIXME: bio_set_dir() */
osd_req_write(or, &obj, i_start, pcol_copy->bio, pcol_copy->length);
ret = exofs_async_op(or, writepages_done, pcol_copy, oi->i_cred); ios->bio = pcol_copy->bio;
ios->offset = pcol_copy->pg_first << PAGE_CACHE_SHIFT;
ios->length = pcol_copy->length;
ios->done = writepages_done;
ios->private = pcol_copy;
ret = exofs_oi_write(oi, ios);
if (unlikely(ret)) { if (unlikely(ret)) {
EXOFS_ERR("write_exec: exofs_async_op() Faild\n"); EXOFS_ERR("write_exec: exofs_oi_write() Faild\n");
goto err; goto err;
} }
atomic_inc(&pcol->sbi->s_curr_pending); atomic_inc(&pcol->sbi->s_curr_pending);
EXOFS_DBGMSG("write_exec(0x%lx, 0x%llx) start=0x%llx length=0x%lx\n", EXOFS_DBGMSG("write_exec(0x%lx, 0x%llx) start=0x%llx length=0x%lx\n",
pcol->inode->i_ino, pcol->pg_first, _LLU(i_start), pcol->inode->i_ino, pcol->pg_first, _LLU(ios->offset),
pcol->length); pcol->length);
/* pages ownership was passed to pcol_copy */ /* pages ownership was passed to pcol_copy */
_pcol_reset(pcol); _pcol_reset(pcol);
@ -542,9 +533,9 @@ static int write_exec(struct page_collect *pcol)
err: err:
_unlock_pcol_pages(pcol, ret, WRITE); _unlock_pcol_pages(pcol, ret, WRITE);
pcol_free(pcol);
kfree(pcol_copy); kfree(pcol_copy);
if (or)
osd_end_request(or);
return ret; return ret;
} }
@ -588,6 +579,9 @@ static int writepage_strip(struct page *page,
if (PageError(page)) if (PageError(page))
ClearPageError(page); ClearPageError(page);
unlock_page(page); unlock_page(page);
EXOFS_DBGMSG("writepage_strip(0x%lx, 0x%lx) "
"outside the limits\n",
inode->i_ino, page->index);
return 0; return 0;
} }
} }
@ -602,6 +596,9 @@ try_again:
ret = write_exec(pcol); ret = write_exec(pcol);
if (unlikely(ret)) if (unlikely(ret))
goto fail; goto fail;
EXOFS_DBGMSG("writepage_strip(0x%lx, 0x%lx) Discontinuity\n",
inode->i_ino, page->index);
goto try_again; goto try_again;
} }
@ -636,6 +633,8 @@ try_again:
return 0; return 0;
fail: fail:
EXOFS_DBGMSG("Error: writepage_strip(0x%lx, 0x%lx)=>%d\n",
inode->i_ino, page->index, ret);
set_bit(AS_EIO, &page->mapping->flags); set_bit(AS_EIO, &page->mapping->flags);
unlock_page(page); unlock_page(page);
return ret; return ret;
@ -654,14 +653,17 @@ static int exofs_writepages(struct address_space *mapping,
wbc->range_end >> PAGE_CACHE_SHIFT; wbc->range_end >> PAGE_CACHE_SHIFT;
if (start || end) if (start || end)
expected_pages = min(end - start + 1, 32L); expected_pages = end - start + 1;
else else
expected_pages = mapping->nrpages; expected_pages = mapping->nrpages;
EXOFS_DBGMSG("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx" if (expected_pages < 32L)
" m->nrpages=%lu start=0x%lx end=0x%lx\n", expected_pages = 32L;
EXOFS_DBGMSG("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx "
"nrpages=%lu start=0x%lx end=0x%lx expected_pages=%ld\n",
mapping->host->i_ino, wbc->range_start, wbc->range_end, mapping->host->i_ino, wbc->range_start, wbc->range_end,
mapping->nrpages, start, end); mapping->nrpages, start, end, expected_pages);
_pcol_init(&pcol, expected_pages, mapping->host); _pcol_init(&pcol, expected_pages, mapping->host);
@ -773,19 +775,28 @@ static int exofs_get_block(struct inode *inode, sector_t iblock,
const struct osd_attr g_attr_logical_length = ATTR_DEF( const struct osd_attr g_attr_logical_length = ATTR_DEF(
OSD_APAGE_OBJECT_INFORMATION, OSD_ATTR_OI_LOGICAL_LENGTH, 8); OSD_APAGE_OBJECT_INFORMATION, OSD_ATTR_OI_LOGICAL_LENGTH, 8);
static int _do_truncate(struct inode *inode)
{
struct exofs_i_info *oi = exofs_i(inode);
loff_t isize = i_size_read(inode);
int ret;
inode->i_mtime = inode->i_ctime = CURRENT_TIME;
nobh_truncate_page(inode->i_mapping, isize, exofs_get_block);
ret = exofs_oi_truncate(oi, (u64)isize);
EXOFS_DBGMSG("(0x%lx) size=0x%llx\n", inode->i_ino, isize);
return ret;
}
/* /*
* Truncate a file to the specified size - all we have to do is set the size * Truncate a file to the specified size - all we have to do is set the size
* attribute. We make sure the object exists first. * attribute. We make sure the object exists first.
*/ */
void exofs_truncate(struct inode *inode) void exofs_truncate(struct inode *inode)
{ {
struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
struct exofs_i_info *oi = exofs_i(inode); struct exofs_i_info *oi = exofs_i(inode);
struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF};
struct osd_request *or;
struct osd_attr attr;
loff_t isize = i_size_read(inode);
__be64 newsize;
int ret; int ret;
if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
@ -795,22 +806,6 @@ void exofs_truncate(struct inode *inode)
return; return;
if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
return; return;
inode->i_mtime = inode->i_ctime = CURRENT_TIME;
nobh_truncate_page(inode->i_mapping, isize, exofs_get_block);
or = osd_start_request(sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("ERROR: exofs_truncate: osd_start_request failed\n");
goto fail;
}
osd_req_set_attributes(or, &obj);
newsize = cpu_to_be64((u64)isize);
attr = g_attr_logical_length;
attr.val_ptr = &newsize;
osd_req_add_set_attr_list(or, &attr, 1);
/* if we are about to truncate an object, and it hasn't been /* if we are about to truncate an object, and it hasn't been
* created yet, wait * created yet, wait
@ -818,8 +813,7 @@ void exofs_truncate(struct inode *inode)
if (unlikely(wait_obj_created(oi))) if (unlikely(wait_obj_created(oi)))
goto fail; goto fail;
ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred); ret = _do_truncate(inode);
osd_end_request(or);
if (ret) if (ret)
goto fail; goto fail;
@ -849,66 +843,57 @@ int exofs_setattr(struct dentry *dentry, struct iattr *iattr)
/* /*
* Read an inode from the OSD, and return it as is. We also return the size * Read an inode from the OSD, and return it as is. We also return the size
* attribute in the 'sanity' argument if we got compiled with debugging turned * attribute in the 'obj_size' argument.
* on.
*/ */
static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi, static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi,
struct exofs_fcb *inode, uint64_t *sanity) struct exofs_fcb *inode, uint64_t *obj_size)
{ {
struct exofs_sb_info *sbi = sb->s_fs_info; struct exofs_sb_info *sbi = sb->s_fs_info;
struct osd_request *or; struct osd_attr attrs[2];
struct osd_attr attr; struct exofs_io_state *ios;
struct osd_obj_id obj = {sbi->s_pid,
oi->vfs_inode.i_ino + EXOFS_OBJ_OFF};
int ret; int ret;
exofs_make_credential(oi->i_cred, &obj); *obj_size = ~0;
ret = exofs_get_io_state(sbi, &ios);
or = osd_start_request(sbi->s_dev, GFP_KERNEL); if (unlikely(ret)) {
if (unlikely(!or)) { EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
EXOFS_ERR("exofs_get_inode: osd_start_request failed.\n"); return ret;
return -ENOMEM;
} }
osd_req_get_attributes(or, &obj);
/* we need the inode attribute */ ios->obj.id = exofs_oi_objno(oi);
osd_req_add_get_attr_list(or, &g_attr_inode_data, 1); exofs_make_credential(oi->i_cred, &ios->obj);
ios->cred = oi->i_cred;
#ifdef EXOFS_DEBUG_OBJ_ISIZE attrs[0] = g_attr_inode_data;
/* we get the size attributes to do a sanity check */ attrs[1] = g_attr_logical_length;
osd_req_add_get_attr_list(or, &g_attr_logical_length, 1); ios->in_attr = attrs;
#endif ios->in_attr_len = ARRAY_SIZE(attrs);
ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred); ret = exofs_sbi_read(ios);
if (ret) if (ret)
goto out; goto out;
attr = g_attr_inode_data; ret = extract_attr_from_ios(ios, &attrs[0]);
ret = extract_attr_from_req(or, &attr);
if (ret) { if (ret) {
EXOFS_ERR("exofs_get_inode: extract_attr_from_req failed\n"); EXOFS_ERR("%s: extract_attr of inode_data failed\n", __func__);
goto out; goto out;
} }
WARN_ON(attrs[0].len != EXOFS_INO_ATTR_SIZE);
memcpy(inode, attrs[0].val_ptr, EXOFS_INO_ATTR_SIZE);
WARN_ON(attr.len != EXOFS_INO_ATTR_SIZE); ret = extract_attr_from_ios(ios, &attrs[1]);
memcpy(inode, attr.val_ptr, EXOFS_INO_ATTR_SIZE);
#ifdef EXOFS_DEBUG_OBJ_ISIZE
attr = g_attr_logical_length;
ret = extract_attr_from_req(or, &attr);
if (ret) { if (ret) {
EXOFS_ERR("ERROR: extract attr from or failed\n"); EXOFS_ERR("%s: extract_attr of logical_length failed\n",
__func__);
goto out; goto out;
} }
*sanity = get_unaligned_be64(attr.val_ptr); *obj_size = get_unaligned_be64(attrs[1].val_ptr);
#endif
out: out:
osd_end_request(or); exofs_put_io_state(ios);
return ret; return ret;
} }
static void __oi_init(struct exofs_i_info *oi) static void __oi_init(struct exofs_i_info *oi)
{ {
init_waitqueue_head(&oi->i_wq); init_waitqueue_head(&oi->i_wq);
@ -922,7 +907,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
struct exofs_i_info *oi; struct exofs_i_info *oi;
struct exofs_fcb fcb; struct exofs_fcb fcb;
struct inode *inode; struct inode *inode;
uint64_t uninitialized_var(sanity); uint64_t obj_size;
int ret; int ret;
inode = iget_locked(sb, ino); inode = iget_locked(sb, ino);
@ -934,7 +919,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
__oi_init(oi); __oi_init(oi);
/* read the inode from the osd */ /* read the inode from the osd */
ret = exofs_get_inode(sb, oi, &fcb, &sanity); ret = exofs_get_inode(sb, oi, &fcb, &obj_size);
if (ret) if (ret)
goto bad_inode; goto bad_inode;
@ -955,13 +940,12 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
inode->i_blkbits = EXOFS_BLKSHIFT; inode->i_blkbits = EXOFS_BLKSHIFT;
inode->i_generation = le32_to_cpu(fcb.i_generation); inode->i_generation = le32_to_cpu(fcb.i_generation);
#ifdef EXOFS_DEBUG_OBJ_ISIZE if ((inode->i_size != obj_size) &&
if ((inode->i_size != sanity) &&
(!exofs_inode_is_fast_symlink(inode))) { (!exofs_inode_is_fast_symlink(inode))) {
EXOFS_ERR("WARNING: Size of inode=%llu != object=%llu\n", EXOFS_ERR("WARNING: Size of inode=%llu != object=%llu\n",
inode->i_size, _LLU(sanity)); inode->i_size, _LLU(obj_size));
/* FIXME: call exofs_inode_recovery() */
} }
#endif
oi->i_dir_start_lookup = 0; oi->i_dir_start_lookup = 0;
@ -1027,23 +1011,30 @@ int __exofs_wait_obj_created(struct exofs_i_info *oi)
* set the obj_created flag so that other methods know that the object exists on * set the obj_created flag so that other methods know that the object exists on
* the OSD. * the OSD.
*/ */
static void create_done(struct osd_request *or, void *p) static void create_done(struct exofs_io_state *ios, void *p)
{ {
struct inode *inode = p; struct inode *inode = p;
struct exofs_i_info *oi = exofs_i(inode); struct exofs_i_info *oi = exofs_i(inode);
struct exofs_sb_info *sbi = inode->i_sb->s_fs_info; struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
int ret; int ret;
ret = exofs_check_ok(or); ret = exofs_check_io(ios, NULL);
osd_end_request(or); exofs_put_io_state(ios);
atomic_dec(&sbi->s_curr_pending); atomic_dec(&sbi->s_curr_pending);
if (unlikely(ret)) { if (unlikely(ret)) {
EXOFS_ERR("object=0x%llx creation faild in pid=0x%llx", EXOFS_ERR("object=0x%llx creation faild in pid=0x%llx",
_LLU(sbi->s_pid), _LLU(inode->i_ino + EXOFS_OBJ_OFF)); _LLU(exofs_oi_objno(oi)), _LLU(sbi->s_pid));
make_bad_inode(inode); /*TODO: When FS is corrupted creation can fail, object already
} else * exist. Get rid of this asynchronous creation, if exist
set_obj_created(oi); * increment the obj counter and try the next object. Until we
* succeed. All these dangling objects will be made into lost
* files by chkfs.exofs
*/
}
set_obj_created(oi);
atomic_dec(&inode->i_count); atomic_dec(&inode->i_count);
wake_up(&oi->i_wq); wake_up(&oi->i_wq);
@ -1058,8 +1049,7 @@ struct inode *exofs_new_inode(struct inode *dir, int mode)
struct inode *inode; struct inode *inode;
struct exofs_i_info *oi; struct exofs_i_info *oi;
struct exofs_sb_info *sbi; struct exofs_sb_info *sbi;
struct osd_request *or; struct exofs_io_state *ios;
struct osd_obj_id obj;
int ret; int ret;
sb = dir->i_sb; sb = dir->i_sb;
@ -1096,28 +1086,28 @@ struct inode *exofs_new_inode(struct inode *dir, int mode)
mark_inode_dirty(inode); mark_inode_dirty(inode);
obj.partition = sbi->s_pid; ret = exofs_get_io_state(sbi, &ios);
obj.id = inode->i_ino + EXOFS_OBJ_OFF; if (unlikely(ret)) {
exofs_make_credential(oi->i_cred, &obj); EXOFS_ERR("exofs_new_inode: exofs_get_io_state failed\n");
return ERR_PTR(ret);
or = osd_start_request(sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("exofs_new_inode: osd_start_request failed\n");
return ERR_PTR(-ENOMEM);
} }
osd_req_create_object(or, &obj); ios->obj.id = exofs_oi_objno(oi);
exofs_make_credential(oi->i_cred, &ios->obj);
/* increment the refcount so that the inode will still be around when we /* increment the refcount so that the inode will still be around when we
* reach the callback * reach the callback
*/ */
atomic_inc(&inode->i_count); atomic_inc(&inode->i_count);
ret = exofs_async_op(or, create_done, inode, oi->i_cred); ios->done = create_done;
ios->private = inode;
ios->cred = oi->i_cred;
ret = exofs_sbi_create(ios);
if (ret) { if (ret) {
atomic_dec(&inode->i_count); atomic_dec(&inode->i_count);
osd_end_request(or); exofs_put_io_state(ios);
return ERR_PTR(-EIO); return ERR_PTR(ret);
} }
atomic_inc(&sbi->s_curr_pending); atomic_inc(&sbi->s_curr_pending);
@ -1135,11 +1125,11 @@ struct updatei_args {
/* /*
* Callback function from exofs_update_inode(). * Callback function from exofs_update_inode().
*/ */
static void updatei_done(struct osd_request *or, void *p) static void updatei_done(struct exofs_io_state *ios, void *p)
{ {
struct updatei_args *args = p; struct updatei_args *args = p;
osd_end_request(or); exofs_put_io_state(ios);
atomic_dec(&args->sbi->s_curr_pending); atomic_dec(&args->sbi->s_curr_pending);
@ -1155,8 +1145,7 @@ static int exofs_update_inode(struct inode *inode, int do_sync)
struct exofs_i_info *oi = exofs_i(inode); struct exofs_i_info *oi = exofs_i(inode);
struct super_block *sb = inode->i_sb; struct super_block *sb = inode->i_sb;
struct exofs_sb_info *sbi = sb->s_fs_info; struct exofs_sb_info *sbi = sb->s_fs_info;
struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF}; struct exofs_io_state *ios;
struct osd_request *or;
struct osd_attr attr; struct osd_attr attr;
struct exofs_fcb *fcb; struct exofs_fcb *fcb;
struct updatei_args *args; struct updatei_args *args;
@ -1193,18 +1182,16 @@ static int exofs_update_inode(struct inode *inode, int do_sync)
} else } else
memcpy(fcb->i_data, oi->i_data, sizeof(fcb->i_data)); memcpy(fcb->i_data, oi->i_data, sizeof(fcb->i_data));
or = osd_start_request(sbi->s_dev, GFP_KERNEL); ret = exofs_get_io_state(sbi, &ios);
if (unlikely(!or)) { if (unlikely(ret)) {
EXOFS_ERR("exofs_update_inode: osd_start_request failed.\n"); EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
ret = -ENOMEM;
goto free_args; goto free_args;
} }
osd_req_set_attributes(or, &obj);
attr = g_attr_inode_data; attr = g_attr_inode_data;
attr.val_ptr = fcb; attr.val_ptr = fcb;
osd_req_add_set_attr_list(or, &attr, 1); ios->out_attr_len = 1;
ios->out_attr = &attr;
if (!obj_created(oi)) { if (!obj_created(oi)) {
EXOFS_DBGMSG("!obj_created\n"); EXOFS_DBGMSG("!obj_created\n");
@ -1213,22 +1200,19 @@ static int exofs_update_inode(struct inode *inode, int do_sync)
EXOFS_DBGMSG("wait_event done\n"); EXOFS_DBGMSG("wait_event done\n");
} }
if (do_sync) { if (!do_sync) {
ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred);
osd_end_request(or);
goto free_args;
} else {
args->sbi = sbi; args->sbi = sbi;
ios->done = updatei_done;
ios->private = args;
}
ret = exofs_async_op(or, updatei_done, args, oi->i_cred); ret = exofs_oi_write(oi, ios);
if (ret) { if (!do_sync && !ret) {
osd_end_request(or);
goto free_args;
}
atomic_inc(&sbi->s_curr_pending); atomic_inc(&sbi->s_curr_pending);
goto out; /* deallocation in updatei_done */ goto out; /* deallocation in updatei_done */
} }
exofs_put_io_state(ios);
free_args: free_args:
kfree(args); kfree(args);
out: out:
@ -1245,11 +1229,12 @@ int exofs_write_inode(struct inode *inode, int wait)
* Callback function from exofs_delete_inode() - don't have much cleaning up to * Callback function from exofs_delete_inode() - don't have much cleaning up to
* do. * do.
*/ */
static void delete_done(struct osd_request *or, void *p) static void delete_done(struct exofs_io_state *ios, void *p)
{ {
struct exofs_sb_info *sbi; struct exofs_sb_info *sbi = p;
osd_end_request(or);
sbi = p; exofs_put_io_state(ios);
atomic_dec(&sbi->s_curr_pending); atomic_dec(&sbi->s_curr_pending);
} }
@ -1263,8 +1248,7 @@ void exofs_delete_inode(struct inode *inode)
struct exofs_i_info *oi = exofs_i(inode); struct exofs_i_info *oi = exofs_i(inode);
struct super_block *sb = inode->i_sb; struct super_block *sb = inode->i_sb;
struct exofs_sb_info *sbi = sb->s_fs_info; struct exofs_sb_info *sbi = sb->s_fs_info;
struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF}; struct exofs_io_state *ios;
struct osd_request *or;
int ret; int ret;
truncate_inode_pages(&inode->i_data, 0); truncate_inode_pages(&inode->i_data, 0);
@ -1281,25 +1265,26 @@ void exofs_delete_inode(struct inode *inode)
clear_inode(inode); clear_inode(inode);
or = osd_start_request(sbi->s_dev, GFP_KERNEL); ret = exofs_get_io_state(sbi, &ios);
if (unlikely(!or)) { if (unlikely(ret)) {
EXOFS_ERR("exofs_delete_inode: osd_start_request failed\n"); EXOFS_ERR("%s: exofs_get_io_state failed\n", __func__);
return; return;
} }
osd_req_remove_object(or, &obj);
/* if we are deleting an obj that hasn't been created yet, wait */ /* if we are deleting an obj that hasn't been created yet, wait */
if (!obj_created(oi)) { if (!obj_created(oi)) {
BUG_ON(!obj_2bcreated(oi)); BUG_ON(!obj_2bcreated(oi));
wait_event(oi->i_wq, obj_created(oi)); wait_event(oi->i_wq, obj_created(oi));
} }
ret = exofs_async_op(or, delete_done, sbi, oi->i_cred); ios->obj.id = exofs_oi_objno(oi);
ios->done = delete_done;
ios->private = sbi;
ios->cred = oi->i_cred;
ret = exofs_sbi_remove(ios);
if (ret) { if (ret) {
EXOFS_ERR( EXOFS_ERR("%s: exofs_sbi_remove failed\n", __func__);
"ERROR: @exofs_delete_inode exofs_async_op failed\n"); exofs_put_io_state(ios);
osd_end_request(or);
return; return;
} }
atomic_inc(&sbi->s_curr_pending); atomic_inc(&sbi->s_curr_pending);

390
fs/exofs/ios.c
View file

@ -23,88 +23,327 @@
*/ */
#include <scsi/scsi_device.h> #include <scsi/scsi_device.h>
#include <scsi/osd_sense.h>
#include "exofs.h" #include "exofs.h"
int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid)
{
struct osd_sense_info osi;
int ret = osd_req_decode_sense(or, &osi);
if (ret) { /* translate to Linux codes */
if (osi.additional_code == scsi_invalid_field_in_cdb) {
if (osi.cdb_field_offset == OSD_CFO_STARTING_BYTE)
ret = -EFAULT;
if (osi.cdb_field_offset == OSD_CFO_OBJECT_ID)
ret = -ENOENT;
else
ret = -EINVAL;
} else if (osi.additional_code == osd_quota_error)
ret = -ENOSPC;
else
ret = -EIO;
}
/* FIXME: should be include in osd_sense_info */
if (in_resid)
*in_resid = or->in.req ? or->in.req->resid_len : 0;
if (out_resid)
*out_resid = or->out.req ? or->out.req->resid_len : 0;
return ret;
}
void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj) void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj)
{ {
osd_sec_init_nosec_doall_caps(cred_a, obj, false, true); osd_sec_init_nosec_doall_caps(cred_a, obj, false, true);
} }
/* int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
* Perform a synchronous OSD operation. u64 offset, void *p, unsigned length)
*/
int exofs_sync_op(struct osd_request *or, int timeout, uint8_t *credential)
{ {
struct osd_request *or = osd_start_request(od, GFP_KERNEL);
/* struct osd_sense_info osi = {.key = 0};*/
int ret; int ret;
or->timeout = timeout; if (unlikely(!or)) {
ret = osd_finalize_request(or, 0, credential, NULL); EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__);
if (ret) { return -ENOMEM;
}
ret = osd_req_read_kern(or, obj, offset, p, length);
if (unlikely(ret)) {
EXOFS_DBGMSG("%s: osd_req_read_kern failed.\n", __func__);
goto out;
}
ret = osd_finalize_request(or, 0, cred, NULL);
if (unlikely(ret)) {
EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret); EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret);
return ret; goto out;
} }
ret = osd_execute_request(or); ret = osd_execute_request(or);
if (unlikely(ret))
if (ret)
EXOFS_DBGMSG("osd_execute_request() => %d\n", ret); EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
/* osd_req_decode_sense(or, ret); */ /* osd_req_decode_sense(or, ret); */
out:
osd_end_request(or);
return ret; return ret;
} }
/* int exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** pios)
* Perform an asynchronous OSD operation.
*/
int exofs_async_op(struct osd_request *or, osd_req_done_fn *async_done,
void *caller_context, u8 *cred)
{ {
int ret; struct exofs_io_state *ios;
ret = osd_finalize_request(or, 0, cred, NULL); /*TODO: Maybe use kmem_cach per sbi of size
if (ret) { * exofs_io_state_size(sbi->s_numdevs)
EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret); */
return ret; ios = kzalloc(exofs_io_state_size(1), GFP_KERNEL);
if (unlikely(!ios)) {
*pios = NULL;
return -ENOMEM;
} }
ret = osd_execute_request_async(or, async_done, caller_context); ios->sbi = sbi;
ios->obj.partition = sbi->s_pid;
*pios = ios;
return 0;
}
if (ret) void exofs_put_io_state(struct exofs_io_state *ios)
EXOFS_DBGMSG("osd_execute_request_async() => %d\n", ret); {
if (ios) {
unsigned i;
for (i = 0; i < ios->numdevs; i++) {
struct exofs_per_dev_state *per_dev = &ios->per_dev[i];
if (per_dev->or)
osd_end_request(per_dev->or);
if (per_dev->bio)
bio_put(per_dev->bio);
}
kfree(ios);
}
}
static void _sync_done(struct exofs_io_state *ios, void *p)
{
struct completion *waiting = p;
complete(waiting);
}
static void _last_io(struct kref *kref)
{
struct exofs_io_state *ios = container_of(
kref, struct exofs_io_state, kref);
ios->done(ios, ios->private);
}
static void _done_io(struct osd_request *or, void *p)
{
struct exofs_io_state *ios = p;
kref_put(&ios->kref, _last_io);
}
static int exofs_io_execute(struct exofs_io_state *ios)
{
DECLARE_COMPLETION_ONSTACK(wait);
bool sync = (ios->done == NULL);
int i, ret;
if (sync) {
ios->done = _sync_done;
ios->private = &wait;
}
for (i = 0; i < ios->numdevs; i++) {
struct osd_request *or = ios->per_dev[i].or;
if (unlikely(!or))
continue;
ret = osd_finalize_request(or, 0, ios->cred, NULL);
if (unlikely(ret)) {
EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n",
ret);
return ret;
}
}
kref_init(&ios->kref);
for (i = 0; i < ios->numdevs; i++) {
struct osd_request *or = ios->per_dev[i].or;
if (unlikely(!or))
continue;
kref_get(&ios->kref);
osd_execute_request_async(or, _done_io, ios);
}
kref_put(&ios->kref, _last_io);
ret = 0;
if (sync) {
wait_for_completion(&wait);
ret = exofs_check_io(ios, NULL);
}
return ret; return ret;
} }
int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr) int exofs_check_io(struct exofs_io_state *ios, u64 *resid)
{
enum osd_err_priority acumulated_osd_err = 0;
int acumulated_lin_err = 0;
int i;
for (i = 0; i < ios->numdevs; i++) {
struct osd_sense_info osi;
int ret = osd_req_decode_sense(ios->per_dev[i].or, &osi);
if (likely(!ret))
continue;
if (unlikely(ret == -EFAULT)) {
EXOFS_DBGMSG("%s: EFAULT Need page clear\n", __func__);
/*FIXME: All the pages in this device range should:
* clear_highpage(page);
*/
}
if (osi.osd_err_pri >= acumulated_osd_err) {
acumulated_osd_err = osi.osd_err_pri;
acumulated_lin_err = ret;
}
}
/* TODO: raid specific residual calculations */
if (resid) {
if (likely(!acumulated_lin_err))
*resid = 0;
else
*resid = ios->length;
}
return acumulated_lin_err;
}
int exofs_sbi_create(struct exofs_io_state *ios)
{
int i, ret;
for (i = 0; i < 1; i++) {
struct osd_request *or;
or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("%s: osd_start_request failed\n", __func__);
ret = -ENOMEM;
goto out;
}
ios->per_dev[i].or = or;
ios->numdevs++;
osd_req_create_object(or, &ios->obj);
}
ret = exofs_io_execute(ios);
out:
return ret;
}
int exofs_sbi_remove(struct exofs_io_state *ios)
{
int i, ret;
for (i = 0; i < 1; i++) {
struct osd_request *or;
or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("%s: osd_start_request failed\n", __func__);
ret = -ENOMEM;
goto out;
}
ios->per_dev[i].or = or;
ios->numdevs++;
osd_req_remove_object(or, &ios->obj);
}
ret = exofs_io_execute(ios);
out:
return ret;
}
int exofs_sbi_write(struct exofs_io_state *ios)
{
int i, ret;
for (i = 0; i < 1; i++) {
struct osd_request *or;
or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("%s: osd_start_request failed\n", __func__);
ret = -ENOMEM;
goto out;
}
ios->per_dev[i].or = or;
ios->numdevs++;
if (ios->bio) {
struct bio *bio;
bio = ios->bio;
osd_req_write(or, &ios->obj, ios->offset, bio,
ios->length);
/* EXOFS_DBGMSG("write sync=%d\n", sync);*/
} else if (ios->kern_buff) {
osd_req_write_kern(or, &ios->obj, ios->offset,
ios->kern_buff, ios->length);
/* EXOFS_DBGMSG("write_kern sync=%d\n", sync);*/
} else {
osd_req_set_attributes(or, &ios->obj);
/* EXOFS_DBGMSG("set_attributes sync=%d\n", sync);*/
}
if (ios->out_attr)
osd_req_add_set_attr_list(or, ios->out_attr,
ios->out_attr_len);
if (ios->in_attr)
osd_req_add_get_attr_list(or, ios->in_attr,
ios->in_attr_len);
}
ret = exofs_io_execute(ios);
out:
return ret;
}
int exofs_sbi_read(struct exofs_io_state *ios)
{
int i, ret;
for (i = 0; i < 1; i++) {
struct osd_request *or;
or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("%s: osd_start_request failed\n", __func__);
ret = -ENOMEM;
goto out;
}
ios->per_dev[i].or = or;
ios->numdevs++;
if (ios->bio) {
osd_req_read(or, &ios->obj, ios->offset, ios->bio,
ios->length);
/* EXOFS_DBGMSG("read sync=%d\n", sync);*/
} else if (ios->kern_buff) {
osd_req_read_kern(or, &ios->obj, ios->offset,
ios->kern_buff, ios->length);
/* EXOFS_DBGMSG("read_kern sync=%d\n", sync);*/
} else {
osd_req_get_attributes(or, &ios->obj);
/* EXOFS_DBGMSG("get_attributes sync=%d\n", sync);*/
}
if (ios->out_attr)
osd_req_add_set_attr_list(or, ios->out_attr,
ios->out_attr_len);
if (ios->in_attr)
osd_req_add_get_attr_list(or, ios->in_attr,
ios->in_attr_len);
}
ret = exofs_io_execute(ios);
out:
return ret;
}
int extract_attr_from_ios(struct exofs_io_state *ios, struct osd_attr *attr)
{ {
struct osd_attr cur_attr = {.attr_page = 0}; /* start with zeros */ struct osd_attr cur_attr = {.attr_page = 0}; /* start with zeros */
void *iter = NULL; void *iter = NULL;
@ -112,7 +351,8 @@ int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr)
do { do {
nelem = 1; nelem = 1;
osd_req_decode_get_attr_list(or, &cur_attr, &nelem, &iter); osd_req_decode_get_attr_list(ios->per_dev[0].or,
&cur_attr, &nelem, &iter);
if ((cur_attr.attr_page == attr->attr_page) && if ((cur_attr.attr_page == attr->attr_page) &&
(cur_attr.attr_id == attr->attr_id)) { (cur_attr.attr_id == attr->attr_id)) {
attr->len = cur_attr.len; attr->len = cur_attr.len;
@ -123,3 +363,43 @@ int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr)
return -EIO; return -EIO;
} }
int exofs_oi_truncate(struct exofs_i_info *oi, u64 size)
{
struct exofs_sb_info *sbi = oi->vfs_inode.i_sb->s_fs_info;
struct exofs_io_state *ios;
struct osd_attr attr;
__be64 newsize;
int i, ret;
if (exofs_get_io_state(sbi, &ios))
return -ENOMEM;
ios->obj.id = exofs_oi_objno(oi);
ios->cred = oi->i_cred;
newsize = cpu_to_be64(size);
attr = g_attr_logical_length;
attr.val_ptr = &newsize;
for (i = 0; i < 1; i++) {
struct osd_request *or;
or = osd_start_request(sbi->s_dev, GFP_KERNEL);
if (unlikely(!or)) {
EXOFS_ERR("%s: osd_start_request failed\n", __func__);
ret = -ENOMEM;
goto out;
}
ios->per_dev[i].or = or;
ios->numdevs++;
osd_req_set_attributes(or, &ios->obj);
osd_req_add_set_attr_list(or, &attr, 1);
}
ret = exofs_io_execute(ios);
out:
exofs_put_io_state(ios);
return ret;
}

120
fs/exofs/super.c
View file

@ -203,49 +203,40 @@ int exofs_sync_fs(struct super_block *sb, int wait)
{ {
struct exofs_sb_info *sbi; struct exofs_sb_info *sbi;
struct exofs_fscb *fscb; struct exofs_fscb *fscb;
struct osd_request *or; struct exofs_io_state *ios;
struct osd_obj_id obj;
int ret = -ENOMEM; int ret = -ENOMEM;
fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL);
if (!fscb) {
EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
return -ENOMEM;
}
lock_super(sb); lock_super(sb);
sbi = sb->s_fs_info; sbi = sb->s_fs_info;
fscb = &sbi->s_fscb;
ret = exofs_get_io_state(sbi, &ios);
if (ret)
goto out;
ios->length = sizeof(*fscb);
memset(fscb, 0, ios->length);
fscb->s_nextid = cpu_to_le64(sbi->s_nextid); fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles); fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
fscb->s_magic = cpu_to_le16(sb->s_magic); fscb->s_magic = cpu_to_le16(sb->s_magic);
fscb->s_newfs = 0; fscb->s_newfs = 0;
or = osd_start_request(sbi->s_dev, GFP_KERNEL); ios->obj.id = EXOFS_SUPER_ID;
if (unlikely(!or)) { ios->offset = 0;
EXOFS_ERR("exofs_write_super: osd_start_request failed.\n"); ios->kern_buff = fscb;
goto out; ios->cred = sbi->s_cred;
}
obj.partition = sbi->s_pid; ret = exofs_sbi_write(ios);
obj.id = EXOFS_SUPER_ID;
ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
if (unlikely(ret)) { if (unlikely(ret)) {
EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n"); EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
goto out;
}
ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
if (unlikely(ret)) {
EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
goto out; goto out;
} }
sb->s_dirt = 0; sb->s_dirt = 0;
out: out:
if (or) EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
osd_end_request(or); exofs_put_io_state(ios);
unlock_super(sb); unlock_super(sb);
kfree(fscb);
return ret; return ret;
} }
@ -302,24 +293,23 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
struct inode *root; struct inode *root;
struct exofs_mountopt *opts = data; struct exofs_mountopt *opts = data;
struct exofs_sb_info *sbi; /*extended info */ struct exofs_sb_info *sbi; /*extended info */
struct osd_dev *od; /* Master device */
struct exofs_fscb fscb; /*on-disk superblock info */ struct exofs_fscb fscb; /*on-disk superblock info */
struct osd_request *or = NULL;
struct osd_obj_id obj; struct osd_obj_id obj;
int ret; int ret;
sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
if (!sbi) if (!sbi)
return -ENOMEM; return -ENOMEM;
sb->s_fs_info = sbi;
/* use mount options to fill superblock */ /* use mount options to fill superblock */
sbi->s_dev = osduld_path_lookup(opts->dev_name); od = osduld_path_lookup(opts->dev_name);
if (IS_ERR(sbi->s_dev)) { if (IS_ERR(od)) {
ret = PTR_ERR(sbi->s_dev); ret = PTR_ERR(od);
sbi->s_dev = NULL;
goto free_sbi; goto free_sbi;
} }
sbi->s_dev = od;
sbi->s_pid = opts->pid; sbi->s_pid = opts->pid;
sbi->s_timeout = opts->timeout; sbi->s_timeout = opts->timeout;
@ -333,35 +323,13 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
sb->s_bdev = NULL; sb->s_bdev = NULL;
sb->s_dev = 0; sb->s_dev = 0;
/* read data from on-disk superblock object */
obj.partition = sbi->s_pid; obj.partition = sbi->s_pid;
obj.id = EXOFS_SUPER_ID; obj.id = EXOFS_SUPER_ID;
exofs_make_credential(sbi->s_cred, &obj); exofs_make_credential(sbi->s_cred, &obj);
or = osd_start_request(sbi->s_dev, GFP_KERNEL); ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
if (unlikely(!or)) { if (unlikely(ret))
if (!silent)
EXOFS_ERR(
"exofs_fill_super: osd_start_request failed.\n");
ret = -ENOMEM;
goto free_sbi; goto free_sbi;
}
ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
if (unlikely(ret)) {
if (!silent)
EXOFS_ERR(
"exofs_fill_super: osd_req_read_kern failed.\n");
ret = -ENOMEM;
goto free_sbi;
}
ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
if (unlikely(ret)) {
if (!silent)
EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
ret = -EIO;
goto free_sbi;
}
sb->s_magic = le16_to_cpu(fscb.s_magic); sb->s_magic = le16_to_cpu(fscb.s_magic);
sbi->s_nextid = le64_to_cpu(fscb.s_nextid); sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
@ -380,6 +348,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
spin_lock_init(&sbi->s_next_gen_lock); spin_lock_init(&sbi->s_next_gen_lock);
/* set up operation vectors */ /* set up operation vectors */
sb->s_fs_info = sbi;
sb->s_op = &exofs_sops; sb->s_op = &exofs_sops;
sb->s_export_op = &exofs_export_ops; sb->s_export_op = &exofs_export_ops;
root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF); root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
@ -406,16 +375,14 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
} }
_exofs_print_device("Mounting", opts->dev_name, sbi->s_dev, sbi->s_pid); _exofs_print_device("Mounting", opts->dev_name, sbi->s_dev, sbi->s_pid);
ret = 0; return 0;
out:
if (or)
osd_end_request(or);
return ret;
free_sbi: free_sbi:
EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
opts->dev_name, sbi->s_pid, ret);
osduld_put_device(sbi->s_dev); /* NULL safe */ osduld_put_device(sbi->s_dev); /* NULL safe */
kfree(sbi); kfree(sbi);
goto out; return ret;
} }
/* /*
@ -444,7 +411,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
{ {
struct super_block *sb = dentry->d_sb; struct super_block *sb = dentry->d_sb;
struct exofs_sb_info *sbi = sb->s_fs_info; struct exofs_sb_info *sbi = sb->s_fs_info;
struct osd_obj_id obj = {sbi->s_pid, 0}; struct exofs_io_state *ios;
struct osd_attr attrs[] = { struct osd_attr attrs[] = {
ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS, ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)), OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
@ -453,26 +420,25 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
}; };
uint64_t capacity = ULLONG_MAX; uint64_t capacity = ULLONG_MAX;
uint64_t used = ULLONG_MAX; uint64_t used = ULLONG_MAX;
struct osd_request *or;
uint8_t cred_a[OSD_CAP_LEN]; uint8_t cred_a[OSD_CAP_LEN];
int ret; int ret;
/* get used/capacity attributes */ ret = exofs_get_io_state(sbi, &ios);
exofs_make_credential(cred_a, &obj); if (ret) {
EXOFS_DBGMSG("exofs_get_io_state failed.\n");
or = osd_start_request(sbi->s_dev, GFP_KERNEL); return ret;
if (unlikely(!or)) {
EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
return -ENOMEM;
} }
osd_req_get_attributes(or, &obj); exofs_make_credential(cred_a, &ios->obj);
osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs)); ios->cred = sbi->s_cred;
ret = exofs_sync_op(or, sbi->s_timeout, cred_a); ios->in_attr = attrs;
ios->in_attr_len = ARRAY_SIZE(attrs);
ret = exofs_sbi_read(ios);
if (unlikely(ret)) if (unlikely(ret))
goto out; goto out;
ret = extract_attr_from_req(or, &attrs[0]); ret = extract_attr_from_ios(ios, &attrs[0]);
if (likely(!ret)) { if (likely(!ret)) {
capacity = get_unaligned_be64(attrs[0].val_ptr); capacity = get_unaligned_be64(attrs[0].val_ptr);
if (unlikely(!capacity)) if (unlikely(!capacity))
@ -480,7 +446,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
} else } else
EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n"); EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
ret = extract_attr_from_req(or, &attrs[1]); ret = extract_attr_from_ios(ios, &attrs[1]);
if (likely(!ret)) if (likely(!ret))
used = get_unaligned_be64(attrs[1].val_ptr); used = get_unaligned_be64(attrs[1].val_ptr);
else else
@ -497,7 +463,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
buf->f_namelen = EXOFS_NAME_LEN; buf->f_namelen = EXOFS_NAME_LEN;
out: out:
osd_end_request(or); exofs_put_io_state(ios);
return ret; return ret;
} }