# Kbuild - Gets included from the Kernels Makefile and build system
#
-exofs-y := ios.o inode.o file.o symlink.o namei.o dir.o super.o
+# ore module library
+obj-$(CONFIG_ORE) += ore.o
+
+exofs-y := inode.o file.o symlink.o namei.o dir.o super.o
obj-$(CONFIG_EXOFS_FS) += exofs.o
+config ORE
+ tristate
+
config EXOFS_FS
tristate "exofs: OSD based file system support"
depends on SCSI_OSD_ULD
+ select ORE
help
EXOFS is a file system that uses an OSD storage device,
as its backing storage.
#include <linux/fs.h>
#include <linux/time.h>
#include <linux/backing-dev.h>
-#include "common.h"
+#include <scsi/osd_ore.h>
-/* FIXME: Remove once pnfs hits mainline
- * #include <linux/exportfs/pnfs_osd_xdr.h>
- */
-#include "pnfs.h"
+#include "common.h"
#define EXOFS_ERR(fmt, a...) printk(KERN_ERR "exofs: " fmt, ##a)
/* u64 has problems with printk this will cast it to unsigned long long */
#define _LLU(x) (unsigned long long)(x)
-struct exofs_layout {
- osd_id s_pid; /* partition ID of file system*/
-
- /* Our way of looking at the data_map */
- unsigned stripe_unit;
- unsigned mirrors_p1;
-
- unsigned group_width;
- u64 group_depth;
- unsigned group_count;
-
- enum exofs_inode_layout_gen_functions lay_func;
-
- unsigned s_numdevs; /* Num of devices in array */
- struct osd_dev *s_ods[0]; /* Variable length */
-};
-
/*
* our extension to the in-memory superblock
*/
struct exofs_sb_info {
+ struct backing_dev_info bdi; /* register our bdi with VFS */
struct exofs_sb_stats s_ess; /* Written often, pre-allocate*/
int s_timeout; /* timeout for OSD operations */
uint64_t s_nextid; /* highest object ID used */
spinlock_t s_next_gen_lock; /* spinlock for gen # update */
u32 s_next_generation; /* next gen # to use */
atomic_t s_curr_pending; /* number of pending commands */
- uint8_t s_cred[OSD_CAP_LEN]; /* credential for the fscb */
- struct backing_dev_info bdi; /* register our bdi with VFS */
struct pnfs_osd_data_map data_map; /* Default raid to use
* FIXME: Needed ?
*/
-/* struct exofs_layout dir_layout;*/ /* Default dir layout */
- struct exofs_layout layout; /* Default files layout,
- * contains the variable osd_dev
- * array. Keep last */
+ struct ore_layout layout; /* Default files layout */
+ struct ore_comp one_comp; /* id & cred of partition id=0*/
+ struct ore_components comps; /* comps for the partition */
struct osd_dev *_min_one_dev[1]; /* Place holder for one dev */
};
uint32_t i_data[EXOFS_IDATA];/*short symlink names and device #s*/
uint32_t i_dir_start_lookup; /* which page to start lookup */
uint64_t i_commit_size; /* the object's written length */
- uint8_t i_cred[OSD_CAP_LEN];/* all-powerful credential */
+ struct ore_comp one_comp; /* same component for all devices */
+ struct ore_components comps; /* inode view of the device table */
};
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_layout *layout;
- struct osd_obj_id obj;
- u8 *cred;
-
- /* Global read/write IO*/
- loff_t offset;
- unsigned long length;
- void *kern_buff;
-
- struct page **pages;
- unsigned nr_pages;
- unsigned pgbase;
- unsigned pages_consumed;
-
- /* 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;
- loff_t offset;
- unsigned length;
- unsigned dev;
- } 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
*/
return container_of(inode, struct exofs_i_info, vfs_inode);
}
-/*
- * Given a layout, object_number and stripe_index return the associated global
- * dev_index
- */
-unsigned exofs_layout_od_id(struct exofs_layout *layout,
- osd_id obj_no, unsigned layout_index);
/*
* Maximum count of links to a file
*/
* 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_layout *layout,
- 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 */
-unsigned exofs_max_io_pages(struct exofs_layout *layout,
+unsigned exofs_max_io_pages(struct ore_layout *layout,
unsigned expected_pages);
int exofs_setattr(struct dentry *, struct iattr *);
int exofs_write_begin(struct file *file, struct address_space *mapping,
struct inode *);
/* super.c */
+void exofs_make_credential(u8 cred_a[OSD_CAP_LEN],
+ const struct osd_obj_id *obj);
int exofs_sbi_write_stats(struct exofs_sb_info *sbi);
/*********************
/* inode.c */
extern const struct address_space_operations exofs_aops;
-extern const struct osd_attr g_attr_logical_length;
/* namei.c */
extern const struct inode_operations exofs_dir_inode_operations;
extern const struct inode_operations exofs_symlink_inode_operations;
extern const struct inode_operations exofs_fast_symlink_inode_operations;
+/* exofs_init_comps will initialize an ore_components device array
+ * pointing to a single ore_comp struct, and a round-robin view
+ * of the device table.
+ * The first device of each inode is the [inode->ino % num_devices]
+ * and the rest of the devices sequentially following where the
+ * first device is after the last device.
+ * It is assumed that the global device array at @sbi is twice
+ * bigger and that the device table repeats twice.
+ * See: exofs_read_lookup_dev_table()
+ */
+static inline void exofs_init_comps(struct ore_components *comps,
+ struct ore_comp *one_comp,
+ struct exofs_sb_info *sbi, osd_id oid)
+{
+ unsigned dev_mod = (unsigned)oid, first_dev;
+
+ one_comp->obj.partition = sbi->one_comp.obj.partition;
+ one_comp->obj.id = oid;
+ exofs_make_credential(one_comp->cred, &one_comp->obj);
+
+ comps->numdevs = sbi->comps.numdevs;
+ comps->single_comp = EC_SINGLE_COMP;
+ comps->comps = one_comp;
+
+ /* Round robin device view of the table */
+ first_dev = (dev_mod * sbi->layout.mirrors_p1) % sbi->comps.numdevs;
+ comps->ods = sbi->comps.ods + first_dev;
+}
+
#endif
PAGE_SIZE / sizeof(struct page *),
};
-unsigned exofs_max_io_pages(struct exofs_layout *layout,
+unsigned exofs_max_io_pages(struct ore_layout *layout,
unsigned expected_pages)
{
unsigned pages = min_t(unsigned, expected_pages, MAX_PAGES_KMALLOC);
struct exofs_sb_info *sbi;
struct inode *inode;
unsigned expected_pages;
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
struct page **pages;
unsigned alloc_pages;
{
unsigned pages;
- if (!pcol->ios) { /* First time allocate io_state */
- int ret = exofs_get_io_state(&pcol->sbi->layout, &pcol->ios);
-
- if (ret)
- return ret;
- }
-
/* TODO: easily support bio chaining */
pages = exofs_max_io_pages(&pcol->sbi->layout, pcol->expected_pages);
pcol->pages = NULL;
if (pcol->ios) {
- exofs_put_io_state(pcol->ios);
+ ore_put_io_state(pcol->ios);
pcol->ios = NULL;
}
}
u64 resid;
u64 good_bytes;
u64 length = 0;
- int ret = exofs_check_io(pcol->ios, &resid);
+ int ret = ore_check_io(pcol->ios, &resid);
if (likely(!ret))
good_bytes = pcol->length;
}
/* callback of async reads */
-static void readpages_done(struct exofs_io_state *ios, void *p)
+static void readpages_done(struct ore_io_state *ios, void *p)
{
struct page_collect *pcol = p;
static int read_exec(struct page_collect *pcol)
{
struct exofs_i_info *oi = exofs_i(pcol->inode);
- struct exofs_io_state *ios = pcol->ios;
+ struct ore_io_state *ios;
struct page_collect *pcol_copy = NULL;
int ret;
if (!pcol->pages)
return 0;
+ if (!pcol->ios) {
+ int ret = ore_get_rw_state(&pcol->sbi->layout, &oi->comps, true,
+ pcol->pg_first << PAGE_CACHE_SHIFT,
+ pcol->length, &pcol->ios);
+
+ if (ret)
+ return ret;
+ }
+
+ ios = pcol->ios;
ios->pages = pcol->pages;
ios->nr_pages = pcol->nr_pages;
- ios->length = pcol->length;
- ios->offset = pcol->pg_first << PAGE_CACHE_SHIFT;
if (pcol->read_4_write) {
- exofs_oi_read(oi, pcol->ios);
+ ore_read(pcol->ios);
return __readpages_done(pcol);
}
*pcol_copy = *pcol;
ios->done = readpages_done;
ios->private = pcol_copy;
- ret = exofs_oi_read(oi, ios);
+ ret = ore_read(ios);
if (unlikely(ret))
goto err;
atomic_inc(&pcol->sbi->s_curr_pending);
EXOFS_DBGMSG2("read_exec obj=0x%llx start=0x%llx length=0x%lx\n",
- ios->obj.id, _LLU(ios->offset), pcol->length);
+ oi->one_comp.obj.id, _LLU(ios->offset), pcol->length);
/* pages ownership was passed to pcol_copy */
_pcol_reset(pcol);
}
/* Callback for osd_write. All writes are asynchronous */
-static void writepages_done(struct exofs_io_state *ios, void *p)
+static void writepages_done(struct ore_io_state *ios, void *p)
{
struct page_collect *pcol = p;
int i;
u64 resid;
u64 good_bytes;
u64 length = 0;
- int ret = exofs_check_io(ios, &resid);
+ int ret = ore_check_io(ios, &resid);
atomic_dec(&pcol->sbi->s_curr_pending);
static int write_exec(struct page_collect *pcol)
{
struct exofs_i_info *oi = exofs_i(pcol->inode);
- struct exofs_io_state *ios = pcol->ios;
+ struct ore_io_state *ios;
struct page_collect *pcol_copy = NULL;
int ret;
if (!pcol->pages)
return 0;
+ BUG_ON(pcol->ios);
+ ret = ore_get_rw_state(&pcol->sbi->layout, &oi->comps, false,
+ pcol->pg_first << PAGE_CACHE_SHIFT,
+ pcol->length, &pcol->ios);
+
+ if (unlikely(ret))
+ goto err;
+
pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL);
if (!pcol_copy) {
EXOFS_ERR("write_exec: Failed to kmalloc(pcol)\n");
*pcol_copy = *pcol;
+ ios = pcol->ios;
ios->pages = pcol_copy->pages;
ios->nr_pages = pcol_copy->nr_pages;
- 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);
+ ret = ore_write(ios);
if (unlikely(ret)) {
- EXOFS_ERR("write_exec: exofs_oi_write() Failed\n");
+ EXOFS_ERR("write_exec: ore_write() Failed\n");
goto err;
}
return S_ISLNK(inode->i_mode) && (oi->i_data[0] != 0);
}
-const struct osd_attr g_attr_logical_length = ATTR_DEF(
- OSD_APAGE_OBJECT_INFORMATION, OSD_ATTR_OI_LOGICAL_LENGTH, 8);
-
static int _do_truncate(struct inode *inode, loff_t newsize)
{
struct exofs_i_info *oi = exofs_i(inode);
+ struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
int ret;
inode->i_mtime = inode->i_ctime = CURRENT_TIME;
- ret = exofs_oi_truncate(oi, (u64)newsize);
+ ret = ore_truncate(&sbi->layout, &oi->comps, (u64)newsize);
if (likely(!ret))
truncate_setsize(inode, newsize);
[1] = g_attr_inode_file_layout,
[2] = g_attr_inode_dir_layout,
};
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
struct exofs_on_disk_inode_layout *layout;
int ret;
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &oi->comps, &ios);
if (unlikely(ret)) {
- EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
+ EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
return ret;
}
- ios->obj.id = exofs_oi_objno(oi);
- exofs_make_credential(oi->i_cred, &ios->obj);
- ios->cred = oi->i_cred;
-
- attrs[1].len = exofs_on_disk_inode_layout_size(sbi->layout.s_numdevs);
- attrs[2].len = exofs_on_disk_inode_layout_size(sbi->layout.s_numdevs);
+ attrs[1].len = exofs_on_disk_inode_layout_size(sbi->comps.numdevs);
+ attrs[2].len = exofs_on_disk_inode_layout_size(sbi->comps.numdevs);
ios->in_attr = attrs;
ios->in_attr_len = ARRAY_SIZE(attrs);
- ret = exofs_sbi_read(ios);
+ ret = ore_read(ios);
if (unlikely(ret)) {
EXOFS_ERR("object(0x%llx) corrupted, return empty file=>%d\n",
- _LLU(ios->obj.id), ret);
+ _LLU(oi->one_comp.obj.id), ret);
memset(inode, 0, sizeof(*inode));
inode->i_mode = 0040000 | (0777 & ~022);
/* If object is lost on target we might as well enable it's
}
out:
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
return ret;
}
return inode;
oi = exofs_i(inode);
__oi_init(oi);
+ exofs_init_comps(&oi->comps, &oi->one_comp, sb->s_fs_info,
+ exofs_oi_objno(oi));
/* read the inode from the osd */
ret = exofs_get_inode(sb, oi, &fcb);
* set the obj_created flag so that other methods know that the object exists on
* the OSD.
*/
-static void create_done(struct exofs_io_state *ios, void *p)
+static void create_done(struct ore_io_state *ios, void *p)
{
struct inode *inode = p;
struct exofs_i_info *oi = exofs_i(inode);
struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
int ret;
- ret = exofs_check_io(ios, NULL);
- exofs_put_io_state(ios);
+ ret = ore_check_io(ios, NULL);
+ ore_put_io_state(ios);
atomic_dec(&sbi->s_curr_pending);
if (unlikely(ret)) {
EXOFS_ERR("object=0x%llx creation failed in pid=0x%llx",
- _LLU(exofs_oi_objno(oi)), _LLU(sbi->layout.s_pid));
+ _LLU(exofs_oi_objno(oi)),
+ _LLU(oi->one_comp.obj.partition));
/*TODO: When FS is corrupted creation can fail, object already
* exist. Get rid of this asynchronous creation, if exist
* increment the obj counter and try the next object. Until we
*/
struct inode *exofs_new_inode(struct inode *dir, int mode)
{
- struct super_block *sb;
+ struct super_block *sb = dir->i_sb;
+ struct exofs_sb_info *sbi = sb->s_fs_info;
struct inode *inode;
struct exofs_i_info *oi;
- struct exofs_sb_info *sbi;
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
int ret;
- sb = dir->i_sb;
inode = new_inode(sb);
if (!inode)
return ERR_PTR(-ENOMEM);
set_obj_2bcreated(oi);
- sbi = sb->s_fs_info;
-
inode->i_mapping->backing_dev_info = sb->s_bdi;
inode_init_owner(inode, dir, mode);
inode->i_ino = sbi->s_nextid++;
spin_unlock(&sbi->s_next_gen_lock);
insert_inode_hash(inode);
+ exofs_init_comps(&oi->comps, &oi->one_comp, sb->s_fs_info,
+ exofs_oi_objno(oi));
exofs_sbi_write_stats(sbi); /* Make sure new sbi->s_nextid is on disk */
mark_inode_dirty(inode);
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &oi->comps, &ios);
if (unlikely(ret)) {
- EXOFS_ERR("exofs_new_inode: exofs_get_io_state failed\n");
+ EXOFS_ERR("exofs_new_inode: ore_get_io_state failed\n");
return ERR_PTR(ret);
}
- ios->obj.id = exofs_oi_objno(oi);
- exofs_make_credential(oi->i_cred, &ios->obj);
-
ios->done = create_done;
ios->private = inode;
- ios->cred = oi->i_cred;
- ret = exofs_sbi_create(ios);
+
+ ret = ore_create(ios);
if (ret) {
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
return ERR_PTR(ret);
}
atomic_inc(&sbi->s_curr_pending);
/*
* Callback function from exofs_update_inode().
*/
-static void updatei_done(struct exofs_io_state *ios, void *p)
+static void updatei_done(struct ore_io_state *ios, void *p)
{
struct updatei_args *args = p;
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
atomic_dec(&args->sbi->s_curr_pending);
struct exofs_i_info *oi = exofs_i(inode);
struct super_block *sb = inode->i_sb;
struct exofs_sb_info *sbi = sb->s_fs_info;
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
struct osd_attr attr;
struct exofs_fcb *fcb;
struct updatei_args *args;
} else
memcpy(fcb->i_data, oi->i_data, sizeof(fcb->i_data));
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &oi->comps, &ios);
if (unlikely(ret)) {
- EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
+ EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
goto free_args;
}
ios->private = args;
}
- ret = exofs_oi_write(oi, ios);
+ ret = ore_write(ios);
if (!do_sync && !ret) {
atomic_inc(&sbi->s_curr_pending);
goto out; /* deallocation in updatei_done */
}
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
free_args:
kfree(args);
out:
* Callback function from exofs_delete_inode() - don't have much cleaning up to
* do.
*/
-static void delete_done(struct exofs_io_state *ios, void *p)
+static void delete_done(struct ore_io_state *ios, void *p)
{
struct exofs_sb_info *sbi = p;
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
atomic_dec(&sbi->s_curr_pending);
}
struct exofs_i_info *oi = exofs_i(inode);
struct super_block *sb = inode->i_sb;
struct exofs_sb_info *sbi = sb->s_fs_info;
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
int ret;
truncate_inode_pages(&inode->i_data, 0);
/* ignore the error, attempt a remove anyway */
/* Now Remove the OSD objects */
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &oi->comps, &ios);
if (unlikely(ret)) {
- EXOFS_ERR("%s: exofs_get_io_state failed\n", __func__);
+ EXOFS_ERR("%s: ore_get_io_state failed\n", __func__);
return;
}
- ios->obj.id = exofs_oi_objno(oi);
ios->done = delete_done;
ios->private = sbi;
- ios->cred = oi->i_cred;
- ret = exofs_sbi_remove(ios);
+
+ ret = ore_remove(ios);
if (ret) {
- EXOFS_ERR("%s: exofs_sbi_remove failed\n", __func__);
- exofs_put_io_state(ios);
+ EXOFS_ERR("%s: ore_remove failed\n", __func__);
+ ore_put_io_state(ios);
return;
}
atomic_inc(&sbi->s_curr_pending);
*/
#include <linux/slab.h>
-#include <scsi/scsi_device.h>
#include <asm/div64.h>
-#include "exofs.h"
+#include <scsi/osd_ore.h>
-#define EXOFS_DBGMSG2(M...) do {} while (0)
-/* #define EXOFS_DBGMSG2 EXOFS_DBGMSG */
+#define ORE_ERR(fmt, a...) printk(KERN_ERR "ore: " fmt, ##a)
-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);
-}
+#ifdef CONFIG_EXOFS_DEBUG
+#define ORE_DBGMSG(fmt, a...) \
+ printk(KERN_NOTICE "ore @%s:%d: " fmt, __func__, __LINE__, ##a)
+#else
+#define ORE_DBGMSG(fmt, a...) \
+ do { if (0) printk(fmt, ##a); } while (0)
+#endif
-int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
- u64 offset, void *p, unsigned length)
-{
- struct osd_request *or = osd_start_request(od, GFP_KERNEL);
-/* struct osd_sense_info osi = {.key = 0};*/
- int ret;
+/* u64 has problems with printk this will cast it to unsigned long long */
+#define _LLU(x) (unsigned long long)(x)
- if (unlikely(!or)) {
- EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__);
- 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;
- }
+#define ORE_DBGMSG2(M...) do {} while (0)
+/* #define ORE_DBGMSG2 ORE_DBGMSG */
- ret = osd_finalize_request(or, 0, cred, NULL);
- if (unlikely(ret)) {
- EXOFS_DBGMSG("Failed to osd_finalize_request() => %d\n", ret);
- goto out;
- }
+MODULE_AUTHOR("Boaz Harrosh <bharrosh@panasas.com>");
+MODULE_DESCRIPTION("Objects Raid Engine ore.ko");
+MODULE_LICENSE("GPL");
- ret = osd_execute_request(or);
- if (unlikely(ret))
- EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
- /* osd_req_decode_sense(or, ret); */
+static u8 *_ios_cred(struct ore_io_state *ios, unsigned index)
+{
+ return ios->comps->comps[index & ios->comps->single_comp].cred;
+}
-out:
- osd_end_request(or);
- return ret;
+static struct osd_obj_id *_ios_obj(struct ore_io_state *ios, unsigned index)
+{
+ return &ios->comps->comps[index & ios->comps->single_comp].obj;
}
-int exofs_get_io_state(struct exofs_layout *layout,
- struct exofs_io_state **pios)
+static struct osd_dev *_ios_od(struct ore_io_state *ios, unsigned index)
{
- struct exofs_io_state *ios;
+ return ios->comps->ods[index];
+}
+
+int ore_get_rw_state(struct ore_layout *layout, struct ore_components *comps,
+ bool is_reading, u64 offset, u64 length,
+ struct ore_io_state **pios)
+{
+ struct ore_io_state *ios;
/*TODO: Maybe use kmem_cach per sbi of size
* exofs_io_state_size(layout->s_numdevs)
*/
- ios = kzalloc(exofs_io_state_size(layout->s_numdevs), GFP_KERNEL);
+ ios = kzalloc(ore_io_state_size(comps->numdevs), GFP_KERNEL);
if (unlikely(!ios)) {
- EXOFS_DBGMSG("Failed kzalloc bytes=%d\n",
- exofs_io_state_size(layout->s_numdevs));
+ ORE_DBGMSG("Failed kzalloc bytes=%d\n",
+ ore_io_state_size(comps->numdevs));
*pios = NULL;
return -ENOMEM;
}
ios->layout = layout;
- ios->obj.partition = layout->s_pid;
+ ios->comps = comps;
+ ios->offset = offset;
+ ios->length = length;
+ ios->reading = is_reading;
+
*pios = ios;
return 0;
}
+EXPORT_SYMBOL(ore_get_rw_state);
+
+int ore_get_io_state(struct ore_layout *layout, struct ore_components *comps,
+ struct ore_io_state **ios)
+{
+ return ore_get_rw_state(layout, comps, true, 0, 0, ios);
+}
+EXPORT_SYMBOL(ore_get_io_state);
-void exofs_put_io_state(struct exofs_io_state *ios)
+void ore_put_io_state(struct ore_io_state *ios)
{
if (ios) {
unsigned i;
for (i = 0; i < ios->numdevs; i++) {
- struct exofs_per_dev_state *per_dev = &ios->per_dev[i];
+ struct ore_per_dev_state *per_dev = &ios->per_dev[i];
if (per_dev->or)
osd_end_request(per_dev->or);
kfree(ios);
}
}
+EXPORT_SYMBOL(ore_put_io_state);
-unsigned exofs_layout_od_id(struct exofs_layout *layout,
- osd_id obj_no, unsigned layout_index)
-{
-/* switch (layout->lay_func) {
- case LAYOUT_MOVING_WINDOW:
- {*/
- unsigned dev_mod = obj_no;
-
- return (layout_index + dev_mod * layout->mirrors_p1) %
- layout->s_numdevs;
-/* }
- case LAYOUT_FUNC_IMPLICT:
- return layout->devs[layout_index];
- }*/
-}
-
-static inline struct osd_dev *exofs_ios_od(struct exofs_io_state *ios,
- unsigned layout_index)
-{
- return ios->layout->s_ods[
- exofs_layout_od_id(ios->layout, ios->obj.id, layout_index)];
-}
-
-static void _sync_done(struct exofs_io_state *ios, void *p)
+static void _sync_done(struct ore_io_state *ios, void *p)
{
struct completion *waiting = p;
static void _last_io(struct kref *kref)
{
- struct exofs_io_state *ios = container_of(
- kref, struct exofs_io_state, kref);
+ struct ore_io_state *ios = container_of(
+ kref, struct ore_io_state, kref);
ios->done(ios, ios->private);
}
static void _done_io(struct osd_request *or, void *p)
{
- struct exofs_io_state *ios = p;
+ struct ore_io_state *ios = p;
kref_put(&ios->kref, _last_io);
}
-static int exofs_io_execute(struct exofs_io_state *ios)
+static int ore_io_execute(struct ore_io_state *ios)
{
DECLARE_COMPLETION_ONSTACK(wait);
bool sync = (ios->done == NULL);
if (unlikely(!or))
continue;
- ret = osd_finalize_request(or, 0, ios->cred, NULL);
+ ret = osd_finalize_request(or, 0, _ios_cred(ios, i), NULL);
if (unlikely(ret)) {
- EXOFS_DBGMSG("Failed to osd_finalize_request() => %d\n",
+ ORE_DBGMSG("Failed to osd_finalize_request() => %d\n",
ret);
return ret;
}
if (sync) {
wait_for_completion(&wait);
- ret = exofs_check_io(ios, NULL);
+ ret = ore_check_io(ios, NULL);
}
return ret;
}
}
}
-int exofs_check_io(struct exofs_io_state *ios, u64 *resid)
+int ore_check_io(struct ore_io_state *ios, u64 *resid)
{
enum osd_err_priority acumulated_osd_err = 0;
int acumulated_lin_err = 0;
if (OSD_ERR_PRI_CLEAR_PAGES == osi.osd_err_pri) {
/* start read offset passed endof file */
_clear_bio(ios->per_dev[i].bio);
- EXOFS_DBGMSG("start read offset passed end of file "
+ ORE_DBGMSG("start read offset passed end of file "
"offset=0x%llx, length=0x%llx\n",
_LLU(ios->per_dev[i].offset),
_LLU(ios->per_dev[i].length));
return acumulated_lin_err;
}
+EXPORT_SYMBOL(ore_check_io);
/*
* L - logical offset into the file
struct _striping_info {
u64 obj_offset;
u64 group_length;
+ u64 M; /* for truncate */
unsigned dev;
unsigned unit_off;
};
-static void _calc_stripe_info(struct exofs_io_state *ios, u64 file_offset,
+static void _calc_stripe_info(struct ore_layout *layout, u64 file_offset,
struct _striping_info *si)
{
- u32 stripe_unit = ios->layout->stripe_unit;
- u32 group_width = ios->layout->group_width;
- u64 group_depth = ios->layout->group_depth;
+ u32 stripe_unit = layout->stripe_unit;
+ u32 group_width = layout->group_width;
+ u64 group_depth = layout->group_depth;
u32 U = stripe_unit * group_width;
u64 T = U * group_depth;
- u64 S = T * ios->layout->group_count;
+ u64 S = T * layout->group_count;
u64 M = div64_u64(file_offset, S);
/*
/* "H - (N * U)" is just "H % U" so it's bound to u32 */
si->dev = (u32)(H - (N * U)) / stripe_unit + G * group_width;
- si->dev *= ios->layout->mirrors_p1;
+ si->dev *= layout->mirrors_p1;
div_u64_rem(file_offset, stripe_unit, &si->unit_off);
(M * group_depth * stripe_unit);
si->group_length = T - H;
+ si->M = M;
}
-static int _add_stripe_unit(struct exofs_io_state *ios, unsigned *cur_pg,
- unsigned pgbase, struct exofs_per_dev_state *per_dev,
+static int _add_stripe_unit(struct ore_io_state *ios, unsigned *cur_pg,
+ unsigned pgbase, struct ore_per_dev_state *per_dev,
int cur_len)
{
unsigned pg = *cur_pg;
struct request_queue *q =
- osd_request_queue(exofs_ios_od(ios, per_dev->dev));
+ osd_request_queue(_ios_od(ios, per_dev->dev));
per_dev->length += cur_len;
per_dev->bio = bio_kmalloc(GFP_KERNEL, bio_size);
if (unlikely(!per_dev->bio)) {
- EXOFS_DBGMSG("Failed to allocate BIO size=%u\n",
+ ORE_DBGMSG("Failed to allocate BIO size=%u\n",
bio_size);
return -ENOMEM;
}
return 0;
}
-static int _prepare_one_group(struct exofs_io_state *ios, u64 length,
+static int _prepare_one_group(struct ore_io_state *ios, u64 length,
struct _striping_info *si)
{
unsigned stripe_unit = ios->layout->stripe_unit;
int ret = 0;
while (length) {
- struct exofs_per_dev_state *per_dev = &ios->per_dev[dev];
+ struct ore_per_dev_state *per_dev = &ios->per_dev[dev];
unsigned cur_len, page_off = 0;
if (!per_dev->length) {
return ret;
}
-static int _prepare_for_striping(struct exofs_io_state *ios)
+static int _prepare_for_striping(struct ore_io_state *ios)
{
u64 length = ios->length;
u64 offset = ios->offset;
if (!ios->pages) {
if (ios->kern_buff) {
- struct exofs_per_dev_state *per_dev = &ios->per_dev[0];
+ struct ore_per_dev_state *per_dev = &ios->per_dev[0];
- _calc_stripe_info(ios, ios->offset, &si);
+ _calc_stripe_info(ios->layout, ios->offset, &si);
per_dev->offset = si.obj_offset;
per_dev->dev = si.dev;
}
while (length) {
- _calc_stripe_info(ios, offset, &si);
+ _calc_stripe_info(ios->layout, offset, &si);
if (length < si.group_length)
si.group_length = length;
return ret;
}
-int exofs_sbi_create(struct exofs_io_state *ios)
+int ore_create(struct ore_io_state *ios)
{
int i, ret;
- for (i = 0; i < ios->layout->s_numdevs; i++) {
+ for (i = 0; i < ios->comps->numdevs; i++) {
struct osd_request *or;
- or = osd_start_request(exofs_ios_od(ios, i), GFP_KERNEL);
+ or = osd_start_request(_ios_od(ios, i), GFP_KERNEL);
if (unlikely(!or)) {
- EXOFS_ERR("%s: osd_start_request failed\n", __func__);
+ ORE_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);
+ osd_req_create_object(or, _ios_obj(ios, i));
}
- ret = exofs_io_execute(ios);
+ ret = ore_io_execute(ios);
out:
return ret;
}
+EXPORT_SYMBOL(ore_create);
-int exofs_sbi_remove(struct exofs_io_state *ios)
+int ore_remove(struct ore_io_state *ios)
{
int i, ret;
- for (i = 0; i < ios->layout->s_numdevs; i++) {
+ for (i = 0; i < ios->comps->numdevs; i++) {
struct osd_request *or;
- or = osd_start_request(exofs_ios_od(ios, i), GFP_KERNEL);
+ or = osd_start_request(_ios_od(ios, i), GFP_KERNEL);
if (unlikely(!or)) {
- EXOFS_ERR("%s: osd_start_request failed\n", __func__);
+ ORE_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);
+ osd_req_remove_object(or, _ios_obj(ios, i));
}
- ret = exofs_io_execute(ios);
+ ret = ore_io_execute(ios);
out:
return ret;
}
+EXPORT_SYMBOL(ore_remove);
-static int _sbi_write_mirror(struct exofs_io_state *ios, int cur_comp)
+static int _write_mirror(struct ore_io_state *ios, int cur_comp)
{
- struct exofs_per_dev_state *master_dev = &ios->per_dev[cur_comp];
+ struct ore_per_dev_state *master_dev = &ios->per_dev[cur_comp];
unsigned dev = ios->per_dev[cur_comp].dev;
unsigned last_comp = cur_comp + ios->layout->mirrors_p1;
int ret = 0;
return 0; /* Just an empty slot */
for (; cur_comp < last_comp; ++cur_comp, ++dev) {
- struct exofs_per_dev_state *per_dev = &ios->per_dev[cur_comp];
+ struct ore_per_dev_state *per_dev = &ios->per_dev[cur_comp];
struct osd_request *or;
- or = osd_start_request(exofs_ios_od(ios, dev), GFP_KERNEL);
+ or = osd_start_request(_ios_od(ios, dev), GFP_KERNEL);
if (unlikely(!or)) {
- EXOFS_ERR("%s: osd_start_request failed\n", __func__);
+ ORE_ERR("%s: osd_start_request failed\n", __func__);
ret = -ENOMEM;
goto out;
}
bio = bio_kmalloc(GFP_KERNEL,
master_dev->bio->bi_max_vecs);
if (unlikely(!bio)) {
- EXOFS_DBGMSG(
+ ORE_DBGMSG(
"Failed to allocate BIO size=%u\n",
master_dev->bio->bi_max_vecs);
ret = -ENOMEM;
bio->bi_rw |= REQ_WRITE;
}
- osd_req_write(or, &ios->obj, per_dev->offset, bio,
- per_dev->length);
- EXOFS_DBGMSG("write(0x%llx) offset=0x%llx "
+ osd_req_write(or, _ios_obj(ios, dev), per_dev->offset,
+ bio, per_dev->length);
+ ORE_DBGMSG("write(0x%llx) offset=0x%llx "
"length=0x%llx dev=%d\n",
- _LLU(ios->obj.id), _LLU(per_dev->offset),
+ _LLU(_ios_obj(ios, dev)->id),
+ _LLU(per_dev->offset),
_LLU(per_dev->length), dev);
} else if (ios->kern_buff) {
- ret = osd_req_write_kern(or, &ios->obj, per_dev->offset,
- ios->kern_buff, ios->length);
+ ret = osd_req_write_kern(or, _ios_obj(ios, dev),
+ per_dev->offset,
+ ios->kern_buff, ios->length);
if (unlikely(ret))
goto out;
- EXOFS_DBGMSG2("write_kern(0x%llx) offset=0x%llx "
+ ORE_DBGMSG2("write_kern(0x%llx) offset=0x%llx "
"length=0x%llx dev=%d\n",
- _LLU(ios->obj.id), _LLU(per_dev->offset),
+ _LLU(_ios_obj(ios, dev)->id),
+ _LLU(per_dev->offset),
_LLU(ios->length), dev);
} else {
- osd_req_set_attributes(or, &ios->obj);
- EXOFS_DBGMSG2("obj(0x%llx) set_attributes=%d dev=%d\n",
- _LLU(ios->obj.id), ios->out_attr_len, dev);
+ osd_req_set_attributes(or, _ios_obj(ios, dev));
+ ORE_DBGMSG2("obj(0x%llx) set_attributes=%d dev=%d\n",
+ _LLU(_ios_obj(ios, dev)->id),
+ ios->out_attr_len, dev);
}
if (ios->out_attr)
return ret;
}
-int exofs_sbi_write(struct exofs_io_state *ios)
+int ore_write(struct ore_io_state *ios)
{
int i;
int ret;
return ret;
for (i = 0; i < ios->numdevs; i += ios->layout->mirrors_p1) {
- ret = _sbi_write_mirror(ios, i);
+ ret = _write_mirror(ios, i);
if (unlikely(ret))
return ret;
}
- ret = exofs_io_execute(ios);
+ ret = ore_io_execute(ios);
return ret;
}
+EXPORT_SYMBOL(ore_write);
-static int _sbi_read_mirror(struct exofs_io_state *ios, unsigned cur_comp)
+static int _read_mirror(struct ore_io_state *ios, unsigned cur_comp)
{
struct osd_request *or;
- struct exofs_per_dev_state *per_dev = &ios->per_dev[cur_comp];
- unsigned first_dev = (unsigned)ios->obj.id;
+ struct ore_per_dev_state *per_dev = &ios->per_dev[cur_comp];
+ struct osd_obj_id *obj = _ios_obj(ios, cur_comp);
+ unsigned first_dev = (unsigned)obj->id;
if (ios->pages && !per_dev->length)
return 0; /* Just an empty slot */
first_dev = per_dev->dev + first_dev % ios->layout->mirrors_p1;
- or = osd_start_request(exofs_ios_od(ios, first_dev), GFP_KERNEL);
+ or = osd_start_request(_ios_od(ios, first_dev), GFP_KERNEL);
if (unlikely(!or)) {
- EXOFS_ERR("%s: osd_start_request failed\n", __func__);
+ ORE_ERR("%s: osd_start_request failed\n", __func__);
return -ENOMEM;
}
per_dev->or = or;
if (ios->pages) {
- osd_req_read(or, &ios->obj, per_dev->offset,
+ osd_req_read(or, obj, per_dev->offset,
per_dev->bio, per_dev->length);
- EXOFS_DBGMSG("read(0x%llx) offset=0x%llx length=0x%llx"
- " dev=%d\n", _LLU(ios->obj.id),
+ ORE_DBGMSG("read(0x%llx) offset=0x%llx length=0x%llx"
+ " dev=%d\n", _LLU(obj->id),
_LLU(per_dev->offset), _LLU(per_dev->length),
first_dev);
} else if (ios->kern_buff) {
- int ret = osd_req_read_kern(or, &ios->obj, per_dev->offset,
+ int ret = osd_req_read_kern(or, obj, per_dev->offset,
ios->kern_buff, ios->length);
- EXOFS_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
+ ORE_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
"length=0x%llx dev=%d ret=>%d\n",
- _LLU(ios->obj.id), _LLU(per_dev->offset),
+ _LLU(obj->id), _LLU(per_dev->offset),
_LLU(ios->length), first_dev, ret);
if (unlikely(ret))
return ret;
} else {
- osd_req_get_attributes(or, &ios->obj);
- EXOFS_DBGMSG2("obj(0x%llx) get_attributes=%d dev=%d\n",
- _LLU(ios->obj.id), ios->in_attr_len, first_dev);
+ osd_req_get_attributes(or, obj);
+ ORE_DBGMSG2("obj(0x%llx) get_attributes=%d dev=%d\n",
+ _LLU(obj->id),
+ ios->in_attr_len, first_dev);
}
if (ios->out_attr)
osd_req_add_set_attr_list(or, ios->out_attr, ios->out_attr_len);
return 0;
}
-int exofs_sbi_read(struct exofs_io_state *ios)
+int ore_read(struct ore_io_state *ios)
{
int i;
int ret;
return ret;
for (i = 0; i < ios->numdevs; i += ios->layout->mirrors_p1) {
- ret = _sbi_read_mirror(ios, i);
+ ret = _read_mirror(ios, i);
if (unlikely(ret))
return ret;
}
- ret = exofs_io_execute(ios);
+ ret = ore_io_execute(ios);
return ret;
}
+EXPORT_SYMBOL(ore_read);
-int extract_attr_from_ios(struct exofs_io_state *ios, struct osd_attr *attr)
+int extract_attr_from_ios(struct ore_io_state *ios, struct osd_attr *attr)
{
struct osd_attr cur_attr = {.attr_page = 0}; /* start with zeros */
void *iter = NULL;
return -EIO;
}
+EXPORT_SYMBOL(extract_attr_from_ios);
-static int _truncate_mirrors(struct exofs_io_state *ios, unsigned cur_comp,
+static int _truncate_mirrors(struct ore_io_state *ios, unsigned cur_comp,
struct osd_attr *attr)
{
int last_comp = cur_comp + ios->layout->mirrors_p1;
for (; cur_comp < last_comp; ++cur_comp) {
- struct exofs_per_dev_state *per_dev = &ios->per_dev[cur_comp];
+ struct ore_per_dev_state *per_dev = &ios->per_dev[cur_comp];
struct osd_request *or;
- or = osd_start_request(exofs_ios_od(ios, cur_comp), GFP_KERNEL);
+ or = osd_start_request(_ios_od(ios, cur_comp), GFP_KERNEL);
if (unlikely(!or)) {
- EXOFS_ERR("%s: osd_start_request failed\n", __func__);
+ ORE_ERR("%s: osd_start_request failed\n", __func__);
return -ENOMEM;
}
per_dev->or = or;
- osd_req_set_attributes(or, &ios->obj);
+ osd_req_set_attributes(or, _ios_obj(ios, cur_comp));
osd_req_add_set_attr_list(or, attr, 1);
}
return 0;
}
-int exofs_oi_truncate(struct exofs_i_info *oi, u64 size)
+struct _trunc_info {
+ struct _striping_info si;
+ u64 prev_group_obj_off;
+ u64 next_group_obj_off;
+
+ unsigned first_group_dev;
+ unsigned nex_group_dev;
+ unsigned max_devs;
+};
+
+void _calc_trunk_info(struct ore_layout *layout, u64 file_offset,
+ struct _trunc_info *ti)
+{
+ unsigned stripe_unit = layout->stripe_unit;
+
+ _calc_stripe_info(layout, file_offset, &ti->si);
+
+ ti->prev_group_obj_off = ti->si.M * stripe_unit;
+ ti->next_group_obj_off = ti->si.M ? (ti->si.M - 1) * stripe_unit : 0;
+
+ ti->first_group_dev = ti->si.dev - (ti->si.dev % layout->group_width);
+ ti->nex_group_dev = ti->first_group_dev + layout->group_width;
+ ti->max_devs = layout->group_width * layout->group_count;
+}
+
+int ore_truncate(struct ore_layout *layout, struct ore_components *comps,
+ u64 size)
{
- struct exofs_sb_info *sbi = oi->vfs_inode.i_sb->s_fs_info;
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
struct exofs_trunc_attr {
struct osd_attr attr;
__be64 newsize;
} *size_attrs;
- struct _striping_info si;
+ struct _trunc_info ti;
int i, ret;
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(layout, comps, &ios);
if (unlikely(ret))
return ret;
- size_attrs = kcalloc(ios->layout->group_width, sizeof(*size_attrs),
+ _calc_trunk_info(ios->layout, size, &ti);
+
+ size_attrs = kcalloc(ti.max_devs, sizeof(*size_attrs),
GFP_KERNEL);
if (unlikely(!size_attrs)) {
ret = -ENOMEM;
goto out;
}
- ios->obj.id = exofs_oi_objno(oi);
- ios->cred = oi->i_cred;
+ ios->numdevs = ios->comps->numdevs;
- ios->numdevs = ios->layout->s_numdevs;
- _calc_stripe_info(ios, size, &si);
-
- for (i = 0; i < ios->layout->group_width; ++i) {
+ for (i = 0; i < ti.max_devs; ++i) {
struct exofs_trunc_attr *size_attr = &size_attrs[i];
u64 obj_size;
- if (i < si.dev)
- obj_size = si.obj_offset +
- ios->layout->stripe_unit - si.unit_off;
- else if (i == si.dev)
- obj_size = si.obj_offset;
- else /* i > si.dev */
- obj_size = si.obj_offset - si.unit_off;
+ if (i < ti.first_group_dev)
+ obj_size = ti.prev_group_obj_off;
+ else if (i >= ti.nex_group_dev)
+ obj_size = ti.next_group_obj_off;
+ else if (i < ti.si.dev) /* dev within this group */
+ obj_size = ti.si.obj_offset +
+ ios->layout->stripe_unit - ti.si.unit_off;
+ else if (i == ti.si.dev)
+ obj_size = ti.si.obj_offset;
+ else /* i > ti.dev */
+ obj_size = ti.si.obj_offset - ti.si.unit_off;
size_attr->newsize = cpu_to_be64(obj_size);
size_attr->attr = g_attr_logical_length;
size_attr->attr.val_ptr = &size_attr->newsize;
+ ORE_DBGMSG("trunc(0x%llx) obj_offset=0x%llx dev=%d\n",
+ _LLU(comps->comps->obj.id), _LLU(obj_size), i);
ret = _truncate_mirrors(ios, i * ios->layout->mirrors_p1,
&size_attr->attr);
if (unlikely(ret))
goto out;
}
- ret = exofs_io_execute(ios);
+ ret = ore_io_execute(ios);
out:
kfree(size_attrs);
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
return ret;
}
+EXPORT_SYMBOL(ore_truncate);
+
+const struct osd_attr g_attr_logical_length = ATTR_DEF(
+ OSD_APAGE_OBJECT_INFORMATION, OSD_ATTR_OI_LOGICAL_LENGTH, 8);
+EXPORT_SYMBOL(g_attr_logical_length);
+++ /dev/null
-/*
- * Copyright (C) 2008, 2009
- * Boaz Harrosh <bharrosh@panasas.com>
- *
- * This file is part of exofs.
- *
- * exofs is free software; you can redistribute it and/or modify it under the
- * terms of the GNU General Public License version 2 as published by the Free
- * Software Foundation.
- *
- */
-
-/* FIXME: Remove this file once pnfs hits mainline */
-
-#ifndef __EXOFS_PNFS_H__
-#define __EXOFS_PNFS_H__
-
-#if ! defined(__PNFS_OSD_XDR_H__)
-
-enum pnfs_iomode {
- IOMODE_READ = 1,
- IOMODE_RW = 2,
- IOMODE_ANY = 3,
-};
-
-/* Layout Structure */
-enum pnfs_osd_raid_algorithm4 {
- PNFS_OSD_RAID_0 = 1,
- PNFS_OSD_RAID_4 = 2,
- PNFS_OSD_RAID_5 = 3,
- PNFS_OSD_RAID_PQ = 4 /* Reed-Solomon P+Q */
-};
-
-struct pnfs_osd_data_map {
- u32 odm_num_comps;
- u64 odm_stripe_unit;
- u32 odm_group_width;
- u32 odm_group_depth;
- u32 odm_mirror_cnt;
- u32 odm_raid_algorithm;
-};
-
-#endif /* ! defined(__PNFS_OSD_XDR_H__) */
-
-#endif /* __EXOFS_PNFS_H__ */
#include "exofs.h"
+#define EXOFS_DBGMSG2(M...) do {} while (0)
+
/******************************************************************************
* MOUNT OPTIONS
*****************************************************************************/
}
/******************************************************************************
- * SUPERBLOCK FUNCTIONS
+ * Some osd helpers
*****************************************************************************/
-static const struct super_operations exofs_sops;
-static const struct export_operations exofs_export_ops;
+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);
+}
+
+static int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
+ u64 offset, void *p, unsigned length)
+{
+ struct osd_request *or = osd_start_request(od, GFP_KERNEL);
+/* struct osd_sense_info osi = {.key = 0};*/
+ int ret;
+
+ if (unlikely(!or)) {
+ EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__);
+ 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("Failed to osd_finalize_request() => %d\n", ret);
+ goto out;
+ }
+
+ ret = osd_execute_request(or);
+ if (unlikely(ret))
+ EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
+ /* osd_req_decode_sense(or, ret); */
+
+out:
+ osd_end_request(or);
+ EXOFS_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
+ "length=0x%llx dev=%p ret=>%d\n",
+ _LLU(obj->id), _LLU(offset), _LLU(length), od, ret);
+ return ret;
+}
static const struct osd_attr g_attr_sb_stats = ATTR_DEF(
EXOFS_APAGE_SB_DATA,
struct osd_attr attrs[] = {
[0] = g_attr_sb_stats,
};
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
int ret;
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &sbi->comps, &ios);
if (unlikely(ret)) {
- EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
+ EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
return ret;
}
- ios->cred = sbi->s_cred;
-
ios->in_attr = attrs;
ios->in_attr_len = ARRAY_SIZE(attrs);
- ret = exofs_sbi_read(ios);
+ ret = ore_read(ios);
if (unlikely(ret)) {
EXOFS_ERR("Error reading super_block stats => %d\n", ret);
goto out;
}
out:
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
return ret;
}
-static void stats_done(struct exofs_io_state *ios, void *p)
+static void stats_done(struct ore_io_state *ios, void *p)
{
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
/* Good thanks nothing to do anymore */
}
struct osd_attr attrs[] = {
[0] = g_attr_sb_stats,
};
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
int ret;
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &sbi->comps, &ios);
if (unlikely(ret)) {
- EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
+ EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
return ret;
}
sbi->s_ess.s_numfiles = cpu_to_le64(sbi->s_numfiles);
attrs[0].val_ptr = &sbi->s_ess;
- ios->cred = sbi->s_cred;
+
ios->done = stats_done;
ios->private = sbi;
ios->out_attr = attrs;
ios->out_attr_len = ARRAY_SIZE(attrs);
- ret = exofs_sbi_write(ios);
+ ret = ore_write(ios);
if (unlikely(ret)) {
- EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
- exofs_put_io_state(ios);
+ EXOFS_ERR("%s: ore_write failed.\n", __func__);
+ ore_put_io_state(ios);
}
return ret;
}
+/******************************************************************************
+ * SUPERBLOCK FUNCTIONS
+ *****************************************************************************/
+static const struct super_operations exofs_sops;
+static const struct export_operations exofs_export_ops;
+
/*
* Write the superblock to the OSD
*/
{
struct exofs_sb_info *sbi;
struct exofs_fscb *fscb;
- struct exofs_io_state *ios;
+ struct ore_comp one_comp;
+ struct ore_components comps;
+ struct ore_io_state *ios;
int ret = -ENOMEM;
fscb = kmalloc(sizeof(*fscb), GFP_KERNEL);
* version). Otherwise the exofs_fscb is read-only from mkfs time. All
* the writeable info is set in exofs_sbi_write_stats() above.
*/
- ret = exofs_get_io_state(&sbi->layout, &ios);
+
+ exofs_init_comps(&comps, &one_comp, sbi, EXOFS_SUPER_ID);
+
+ ret = ore_get_io_state(&sbi->layout, &comps, &ios);
if (unlikely(ret))
goto out;
fscb->s_newfs = 0;
fscb->s_version = EXOFS_FSCB_VER;
- ios->obj.id = EXOFS_SUPER_ID;
ios->offset = 0;
ios->kern_buff = fscb;
- ios->cred = sbi->s_cred;
- ret = exofs_sbi_write(ios);
+ ret = ore_write(ios);
if (unlikely(ret))
- EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
+ EXOFS_ERR("%s: ore_write failed.\n", __func__);
else
sb->s_dirt = 0;
unlock_super(sb);
out:
EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
kfree(fscb);
return ret;
}
void exofs_free_sbi(struct exofs_sb_info *sbi)
{
- while (sbi->layout.s_numdevs) {
- int i = --sbi->layout.s_numdevs;
- struct osd_dev *od = sbi->layout.s_ods[i];
+ while (sbi->comps.numdevs) {
+ int i = --sbi->comps.numdevs;
+ struct osd_dev *od = sbi->comps.ods[i];
if (od) {
- sbi->layout.s_ods[i] = NULL;
+ sbi->comps.ods[i] = NULL;
osduld_put_device(od);
}
}
+ if (sbi->comps.ods != sbi->_min_one_dev)
+ kfree(sbi->comps.ods);
kfree(sbi);
}
msecs_to_jiffies(100));
}
- _exofs_print_device("Unmounting", NULL, sbi->layout.s_ods[0],
- sbi->layout.s_pid);
+ _exofs_print_device("Unmounting", NULL, sbi->comps.ods[0],
+ sbi->one_comp.obj.partition);
bdi_destroy(&sbi->bdi);
exofs_free_sbi(sbi);
return -EINVAL;
}
+ EXOFS_DBGMSG("exofs: layout: "
+ "num_comps=%u stripe_unit=0x%x group_width=%u "
+ "group_depth=0x%llx mirrors_p1=%u raid_algorithm=%u\n",
+ numdevs,
+ sbi->layout.stripe_unit,
+ sbi->layout.group_width,
+ _LLU(sbi->layout.group_depth),
+ sbi->layout.mirrors_p1,
+ sbi->data_map.odm_raid_algorithm);
return 0;
}
-static unsigned __ra_pages(struct exofs_layout *layout)
+static unsigned __ra_pages(struct ore_layout *layout)
{
const unsigned _MIN_RA = 32; /* min 128K read-ahead */
unsigned ra_pages = layout->group_width * layout->stripe_unit /
return !(odi->systemid_len || odi->osdname_len);
}
-static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
+static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
+ struct osd_dev *fscb_od,
unsigned table_count)
{
- struct exofs_sb_info *sbi = *psbi;
- struct osd_dev *fscb_od;
- struct osd_obj_id obj = {.partition = sbi->layout.s_pid,
- .id = EXOFS_DEVTABLE_ID};
+ struct ore_comp comp;
struct exofs_device_table *dt;
unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) +
sizeof(*dt);
return -ENOMEM;
}
- fscb_od = sbi->layout.s_ods[0];
- sbi->layout.s_ods[0] = NULL;
- sbi->layout.s_numdevs = 0;
- ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes);
+ sbi->comps.numdevs = 0;
+
+ comp.obj.partition = sbi->one_comp.obj.partition;
+ comp.obj.id = EXOFS_DEVTABLE_ID;
+ exofs_make_credential(comp.cred, &comp.obj);
+
+ ret = exofs_read_kern(fscb_od, comp.cred, &comp.obj, 0, dt,
+ table_bytes);
if (unlikely(ret)) {
EXOFS_ERR("ERROR: reading device table\n");
goto out;
goto out;
if (likely(numdevs > 1)) {
- unsigned size = numdevs * sizeof(sbi->layout.s_ods[0]);
+ unsigned size = numdevs * sizeof(sbi->comps.ods[0]);
- sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL);
- if (unlikely(!sbi)) {
+ /* Twice bigger table: See exofs_init_comps() and below
+ * comment
+ */
+ sbi->comps.ods = kzalloc(size + size - 1, GFP_KERNEL);
+ if (unlikely(!sbi->comps.ods)) {
+ EXOFS_ERR("ERROR: faild allocating Device array[%d]\n",
+ numdevs);
ret = -ENOMEM;
goto out;
}
- memset(&sbi->layout.s_ods[1], 0,
- size - sizeof(sbi->layout.s_ods[0]));
- *psbi = sbi;
}
for (i = 0; i < numdevs; i++) {
* line. We always keep them in device-table order.
*/
if (fscb_od && osduld_device_same(fscb_od, &odi)) {
- sbi->layout.s_ods[i] = fscb_od;
- ++sbi->layout.s_numdevs;
+ sbi->comps.ods[i] = fscb_od;
+ ++sbi->comps.numdevs;
fscb_od = NULL;
continue;
}
goto out;
}
- sbi->layout.s_ods[i] = od;
- ++sbi->layout.s_numdevs;
+ sbi->comps.ods[i] = od;
+ ++sbi->comps.numdevs;
/* Read the fscb of the other devices to make sure the FS
* partition is there.
*/
- ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb,
+ ret = exofs_read_kern(od, comp.cred, &comp.obj, 0, &fscb,
sizeof(fscb));
if (unlikely(ret)) {
EXOFS_ERR("ERROR: Malformed participating device "
out:
kfree(dt);
- if (unlikely(!ret && fscb_od)) {
- EXOFS_ERR(
- "ERROR: Bad device-table container device not present\n");
- osduld_put_device(fscb_od);
- ret = -EINVAL;
- }
+ if (likely(!ret)) {
+ unsigned numdevs = sbi->comps.numdevs;
+ if (unlikely(fscb_od)) {
+ EXOFS_ERR("ERROR: Bad device-table container device not present\n");
+ osduld_put_device(fscb_od);
+ return -EINVAL;
+ }
+ /* exofs round-robins the device table view according to inode
+ * number. We hold a: twice bigger table hence inodes can point
+ * to any device and have a sequential view of the table
+ * starting at this device. See exofs_init_comps()
+ */
+ for (i = 0; i < numdevs - 1; ++i)
+ sbi->comps.ods[i + numdevs] = sbi->comps.ods[i];
+ }
return ret;
}
struct exofs_sb_info *sbi; /*extended info */
struct osd_dev *od; /* Master device */
struct exofs_fscb fscb; /*on-disk superblock info */
- struct osd_obj_id obj;
+ struct ore_comp comp;
unsigned table_count;
int ret;
if (!sbi)
return -ENOMEM;
- ret = bdi_setup_and_register(&sbi->bdi, "exofs", BDI_CAP_MAP_COPY);
- if (ret)
- goto free_bdi;
-
/* use mount options to fill superblock */
if (opts->is_osdname) {
struct osd_dev_info odi = {.systemid_len = 0};
odi.osdname_len = strlen(opts->dev_name);
odi.osdname = (u8 *)opts->dev_name;
od = osduld_info_lookup(&odi);
+ kfree(opts->dev_name);
+ opts->dev_name = NULL;
} else {
od = osduld_path_lookup(opts->dev_name);
}
sbi->layout.group_width = 1;
sbi->layout.group_depth = -1;
sbi->layout.group_count = 1;
- sbi->layout.s_ods[0] = od;
- sbi->layout.s_numdevs = 1;
- sbi->layout.s_pid = opts->pid;
sbi->s_timeout = opts->timeout;
+ sbi->one_comp.obj.partition = opts->pid;
+ sbi->one_comp.obj.id = 0;
+ exofs_make_credential(sbi->one_comp.cred, &sbi->one_comp.obj);
+ sbi->comps.numdevs = 1;
+ sbi->comps.single_comp = EC_SINGLE_COMP;
+ sbi->comps.comps = &sbi->one_comp;
+ sbi->comps.ods = sbi->_min_one_dev;
+
/* fill in some other data by hand */
memset(sb->s_id, 0, sizeof(sb->s_id));
strcpy(sb->s_id, "exofs");
sb->s_bdev = NULL;
sb->s_dev = 0;
- obj.partition = sbi->layout.s_pid;
- obj.id = EXOFS_SUPER_ID;
- exofs_make_credential(sbi->s_cred, &obj);
+ comp.obj.partition = sbi->one_comp.obj.partition;
+ comp.obj.id = EXOFS_SUPER_ID;
+ exofs_make_credential(comp.cred, &comp.obj);
- ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
+ ret = exofs_read_kern(od, comp.cred, &comp.obj, 0, &fscb, sizeof(fscb));
if (unlikely(ret))
goto free_sbi;
table_count = le64_to_cpu(fscb.s_dev_table_count);
if (table_count) {
- ret = exofs_read_lookup_dev_table(&sbi, table_count);
+ ret = exofs_read_lookup_dev_table(sbi, od, table_count);
if (unlikely(ret))
goto free_sbi;
+ } else {
+ sbi->comps.ods[0] = od;
}
__sbi_read_stats(sbi);
goto free_sbi;
}
- _exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0],
- sbi->layout.s_pid);
- if (opts->is_osdname)
- kfree(opts->dev_name);
+ ret = bdi_setup_and_register(&sbi->bdi, "exofs", BDI_CAP_MAP_COPY);
+ if (ret) {
+ EXOFS_DBGMSG("Failed to bdi_setup_and_register\n");
+ goto free_sbi;
+ }
+
+ _exofs_print_device("Mounting", opts->dev_name, sbi->comps.ods[0],
+ sbi->one_comp.obj.partition);
return 0;
free_sbi:
- bdi_destroy(&sbi->bdi);
-free_bdi:
EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
- opts->dev_name, sbi->layout.s_pid, ret);
+ opts->dev_name, sbi->one_comp.obj.partition, ret);
exofs_free_sbi(sbi);
- if (opts->is_osdname)
- kfree(opts->dev_name);
return ret;
}
{
struct super_block *sb = dentry->d_sb;
struct exofs_sb_info *sbi = sb->s_fs_info;
- struct exofs_io_state *ios;
+ struct ore_io_state *ios;
struct osd_attr attrs[] = {
ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
};
uint64_t capacity = ULLONG_MAX;
uint64_t used = ULLONG_MAX;
- uint8_t cred_a[OSD_CAP_LEN];
int ret;
- ret = exofs_get_io_state(&sbi->layout, &ios);
+ ret = ore_get_io_state(&sbi->layout, &sbi->comps, &ios);
if (ret) {
- EXOFS_DBGMSG("exofs_get_io_state failed.\n");
+ EXOFS_DBGMSG("ore_get_io_state failed.\n");
return ret;
}
- exofs_make_credential(cred_a, &ios->obj);
- ios->cred = sbi->s_cred;
ios->in_attr = attrs;
ios->in_attr_len = ARRAY_SIZE(attrs);
- ret = exofs_sbi_read(ios);
+ ret = ore_read(ios);
if (unlikely(ret))
goto out;
buf->f_namelen = EXOFS_NAME_LEN;
out:
- exofs_put_io_state(ios);
+ ore_put_io_state(ios);
return ret;
}
struct posix_acl * acl_default;
};
+struct nfs4_string {
+ unsigned int len;
+ char *data;
+};
+
#ifdef CONFIG_NFS_V4
typedef u64 clientid4;
struct nfs4_sequence_res seq_res;
};
-struct nfs4_string {
- unsigned int len;
- char *data;
-};
-
#define NFS4_PATHNAME_MAXCOMPONENTS 512
struct nfs4_pathname {
unsigned int ncomponents;
--- /dev/null
+/*
+ * Copyright (C) 2011
+ * Boaz Harrosh <bharrosh@panasas.com>
+ *
+ * Public Declarations of the ORE API
+ *
+ * This file is part of the ORE (Object Raid Engine) library.
+ *
+ * ORE is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation. (GPL v2)
+ *
+ * ORE is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with the ORE; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+#ifndef __ORE_H__
+#define __ORE_H__
+
+#include <scsi/osd_initiator.h>
+#include <scsi/osd_attributes.h>
+#include <scsi/osd_sec.h>
+#include <linux/pnfs_osd_xdr.h>
+
+struct ore_comp {
+ struct osd_obj_id obj;
+ u8 cred[OSD_CAP_LEN];
+};
+
+struct ore_layout {
+ /* Our way of looking at the data_map */
+ unsigned stripe_unit;
+ unsigned mirrors_p1;
+
+ unsigned group_width;
+ u64 group_depth;
+ unsigned group_count;
+};
+
+struct ore_components {
+ unsigned numdevs; /* Num of devices in array */
+ /* If @single_comp == EC_SINGLE_COMP, @comps points to a single
+ * component. else there are @numdevs components
+ */
+ enum EC_COMP_USAGE {
+ EC_SINGLE_COMP = 0, EC_MULTPLE_COMPS = 0xffffffff
+ } single_comp;
+ struct ore_comp *comps;
+ struct osd_dev **ods; /* osd_dev array */
+};
+
+struct ore_io_state;
+typedef void (*ore_io_done_fn)(struct ore_io_state *ios, void *private);
+
+struct ore_io_state {
+ struct kref kref;
+
+ void *private;
+ ore_io_done_fn done;
+
+ struct ore_layout *layout;
+ struct ore_components *comps;
+
+ /* Global read/write IO*/
+ loff_t offset;
+ unsigned long length;
+ void *kern_buff;
+
+ struct page **pages;
+ unsigned nr_pages;
+ unsigned pgbase;
+ unsigned pages_consumed;
+
+ /* Attributes */
+ unsigned in_attr_len;
+ struct osd_attr *in_attr;
+ unsigned out_attr_len;
+ struct osd_attr *out_attr;
+
+ bool reading;
+
+ /* Variable array of size numdevs */
+ unsigned numdevs;
+ struct ore_per_dev_state {
+ struct osd_request *or;
+ struct bio *bio;
+ loff_t offset;
+ unsigned length;
+ unsigned dev;
+ } per_dev[];
+};
+
+static inline unsigned ore_io_state_size(unsigned numdevs)
+{
+ return sizeof(struct ore_io_state) +
+ sizeof(struct ore_per_dev_state) * numdevs;
+}
+
+/* ore.c */
+int ore_get_rw_state(struct ore_layout *layout, struct ore_components *comps,
+ bool is_reading, u64 offset, u64 length,
+ struct ore_io_state **ios);
+int ore_get_io_state(struct ore_layout *layout, struct ore_components *comps,
+ struct ore_io_state **ios);
+void ore_put_io_state(struct ore_io_state *ios);
+
+int ore_check_io(struct ore_io_state *ios, u64 *resid);
+
+int ore_create(struct ore_io_state *ios);
+int ore_remove(struct ore_io_state *ios);
+int ore_write(struct ore_io_state *ios);
+int ore_read(struct ore_io_state *ios);
+int ore_truncate(struct ore_layout *layout, struct ore_components *comps,
+ u64 size);
+
+int extract_attr_from_ios(struct ore_io_state *ios, struct osd_attr *attr);
+
+extern const struct osd_attr g_attr_logical_length;
+
+#endif
git.openpandora.org / pandora-kernel.git / commitdiff