summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorBoaz Harrosh <bharrosh@panasas.com>2010-02-11 13:01:39 +0200
committerBoaz Harrosh <bharrosh@panasas.com>2010-02-28 03:55:53 -0800
commit50a76fd3c352ed2740eba01512efcfceee0703be (patch)
tree425416e068648e225b41327a120d00bbddd16d0e
parentb367e78bd1c7af4c018ce98b1f6d3e001aba895a (diff)
exofs: groups support
* _calc_stripe_info() changes to accommodate for grouping calculations. Returns additional information * old _prepare_pages() becomes _prepare_one_group() which stores pages belonging to one device group. * New _prepare_for_striping iterates on all groups calling _prepare_one_group(). * Enable mounting of groups data_maps (group_width != 0) [QUESTION] what is faster A or B; A. x += stride; x = x % width + first_x; B x += stride if (x < last_x) x = first_x; Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
-rw-r--r--fs/exofs/exofs.h3
-rw-r--r--fs/exofs/ios.c129
-rw-r--r--fs/exofs/super.c46
3 files changed, 141 insertions, 37 deletions
diff --git a/fs/exofs/exofs.h b/fs/exofs/exofs.h
index acfebd36de83..59b8bf2825c7 100644
--- a/fs/exofs/exofs.h
+++ b/fs/exofs/exofs.h
@@ -63,6 +63,8 @@ struct exofs_layout {
unsigned mirrors_p1;
unsigned group_width;
+ u64 group_depth;
+ unsigned group_count;
enum exofs_inode_layout_gen_functions lay_func;
@@ -132,6 +134,7 @@ struct exofs_io_state {
struct page **pages;
unsigned nr_pages;
unsigned pgbase;
+ unsigned pages_consumed;
/* Attributes */
unsigned in_attr_len;
diff --git a/fs/exofs/ios.c b/fs/exofs/ios.c
index d28febdf54ab..5293bc411d17 100644
--- a/fs/exofs/ios.c
+++ b/fs/exofs/ios.c
@@ -262,25 +262,50 @@ int exofs_check_io(struct exofs_io_state *ios, u64 *resid)
/*
* L - logical offset into the file
*
- * U - The number of bytes in a full stripe
+ * U - The number of bytes in a stripe within a group
*
* U = stripe_unit * group_width
*
- * N - The stripe number
+ * T - The number of bytes striped within a group of component objects
+ * (before advancing to the next group)
*
- * N = L / U
+ * T = stripe_unit * group_width * group_depth
+ *
+ * S - The number of bytes striped across all component objects
+ * before the pattern repeats
+ *
+ * S = stripe_unit * group_width * group_depth * group_count
+ *
+ * M - The "major" (i.e., across all components) stripe number
+ *
+ * M = L / S
+ *
+ * G - Counts the groups from the beginning of the major stripe
+ *
+ * G = (L - (M * S)) / T [or (L % S) / T]
+ *
+ * H - The byte offset within the group
+ *
+ * H = (L - (M * S)) % T [or (L % S) % T]
+ *
+ * N - The "minor" (i.e., across the group) stripe number
+ *
+ * N = H / U
*
* C - The component index coresponding to L
*
- * C = (L - (N*U)) / stripe_unit
+ * C = (H - (N * U)) / stripe_unit + G * group_width
+ * [or (L % U) / stripe_unit + G * group_width]
*
* O - The component offset coresponding to L
*
- * (N*stripe_unit)+(L%stripe_unit)
+ * O = L % stripe_unit + N * stripe_unit + M * group_depth * stripe_unit
*/
-
struct _striping_info {
u64 obj_offset;
+ u64 group_length;
+ u64 total_group_length;
+ u64 Major;
unsigned dev;
unsigned unit_off;
};
@@ -290,15 +315,35 @@ static void _calc_stripe_info(struct exofs_io_state *ios, u64 file_offset,
{
u32 stripe_unit = ios->layout->stripe_unit;
u32 group_width = ios->layout->group_width;
+ u64 group_depth = ios->layout->group_depth;
+
u32 U = stripe_unit * group_width;
+ u64 T = U * group_depth;
+ u64 S = T * ios->layout->group_count;
+ u64 M = div64_u64(file_offset, S);
+
+ /*
+ G = (L - (M * S)) / T
+ H = (L - (M * S)) % T
+ */
+ u64 LmodS = file_offset - M * S;
+ u32 G = div64_u64(LmodS, T);
+ u64 H = LmodS - G * T;
+
+ u32 N = div_u64(H, U);
+
+ /* "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;
- u32 LmodU;
- u64 N = div_u64_rem(file_offset, U, &LmodU);
+ div_u64_rem(file_offset, stripe_unit, &si->unit_off);
- si->unit_off = LmodU % stripe_unit;
- si->obj_offset = N * stripe_unit + si->unit_off;
- si->dev = LmodU / stripe_unit;
- si->dev *= ios->layout->mirrors_p1;
+ si->obj_offset = si->unit_off + (N * stripe_unit) +
+ (M * group_depth * stripe_unit);
+
+ si->group_length = T - H;
+ si->total_group_length = T;
+ si->Major = M;
}
static int _add_stripe_unit(struct exofs_io_state *ios, unsigned *cur_pg,
@@ -345,16 +390,17 @@ static int _add_stripe_unit(struct exofs_io_state *ios, unsigned *cur_pg,
return 0;
}
-static int _prepare_pages(struct exofs_io_state *ios,
- struct _striping_info *si)
+static int _prepare_one_group(struct exofs_io_state *ios, u64 length,
+ struct _striping_info *si, unsigned first_comp)
{
- u64 length = ios->length;
unsigned stripe_unit = ios->layout->stripe_unit;
unsigned mirrors_p1 = ios->layout->mirrors_p1;
+ unsigned devs_in_group = ios->layout->group_width * mirrors_p1;
unsigned dev = si->dev;
- unsigned comp = 0;
- unsigned stripes = 0;
- unsigned cur_pg = 0;
+ unsigned first_dev = dev - (dev % devs_in_group);
+ unsigned comp = first_comp + (dev - first_dev);
+ unsigned max_comp = ios->numdevs ? ios->numdevs - mirrors_p1 : 0;
+ unsigned cur_pg = ios->pages_consumed;
int ret = 0;
while (length) {
@@ -377,10 +423,11 @@ static int _prepare_pages(struct exofs_io_state *ios,
cur_len = stripe_unit;
}
- stripes++;
+ if (max_comp < comp)
+ max_comp = comp;
dev += mirrors_p1;
- dev %= ios->layout->s_numdevs;
+ dev = (dev % devs_in_group) + first_dev;
} else {
cur_len = stripe_unit;
}
@@ -393,18 +440,24 @@ static int _prepare_pages(struct exofs_io_state *ios,
goto out;
comp += mirrors_p1;
- comp %= ios->layout->s_numdevs;
+ comp = (comp % devs_in_group) + first_comp;
length -= cur_len;
}
out:
- ios->numdevs = stripes * mirrors_p1;
+ ios->numdevs = max_comp + mirrors_p1;
+ ios->pages_consumed = cur_pg;
return ret;
}
static int _prepare_for_striping(struct exofs_io_state *ios)
{
+ u64 length = ios->length;
struct _striping_info si;
+ unsigned devs_in_group = ios->layout->group_width *
+ ios->layout->mirrors_p1;
+ unsigned first_comp = 0;
+ int ret = 0;
_calc_stripe_info(ios, ios->offset, &si);
@@ -424,7 +477,31 @@ static int _prepare_for_striping(struct exofs_io_state *ios)
return 0;
}
- return _prepare_pages(ios, &si);
+ while (length) {
+ if (length < si.group_length)
+ si.group_length = length;
+
+ ret = _prepare_one_group(ios, si.group_length, &si, first_comp);
+ if (unlikely(ret))
+ goto out;
+
+ length -= si.group_length;
+
+ si.group_length = si.total_group_length;
+ si.unit_off = 0;
+ ++si.Major;
+ si.obj_offset = si.Major * ios->layout->stripe_unit *
+ ios->layout->group_depth;
+
+ si.dev = (si.dev - (si.dev % devs_in_group)) + devs_in_group;
+ si.dev %= ios->layout->s_numdevs;
+
+ first_comp += devs_in_group;
+ first_comp %= ios->layout->s_numdevs;
+ }
+
+out:
+ return ret;
}
int exofs_sbi_create(struct exofs_io_state *ios)
@@ -482,6 +559,9 @@ static int _sbi_write_mirror(struct exofs_io_state *ios, int cur_comp)
unsigned last_comp = cur_comp + ios->layout->mirrors_p1;
int ret = 0;
+ if (ios->pages && !master_dev->length)
+ 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 osd_request *or;
@@ -580,6 +660,9 @@ static int _sbi_read_mirror(struct exofs_io_state *ios, unsigned cur_comp)
struct exofs_per_dev_state *per_dev = &ios->per_dev[cur_comp];
unsigned first_dev = (unsigned)ios->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);
if (unlikely(!or)) {
diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 8f4e4b37a578..6cf5e4e84d61 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -323,11 +323,7 @@ static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
sbi->data_map.odm_raid_algorithm =
le32_to_cpu(dt->dt_data_map.cb_raid_algorithm);
-/* FIXME: Only raid0 !group_width/depth for now. if not so, do not mount */
- if (sbi->data_map.odm_group_width || sbi->data_map.odm_group_depth) {
- EXOFS_ERR("Group width/depth not supported\n");
- return -EINVAL;
- }
+/* FIXME: Only raid0 for now. if not so, do not mount */
if (sbi->data_map.odm_num_comps != numdevs) {
EXOFS_ERR("odm_num_comps(%u) != numdevs(%u)\n",
sbi->data_map.odm_num_comps, numdevs);
@@ -343,14 +339,6 @@ static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
return -EINVAL;
}
- stripe_length = sbi->data_map.odm_stripe_unit *
- (numdevs / (sbi->data_map.odm_mirror_cnt + 1));
- if (stripe_length >= (1ULL << 32)) {
- EXOFS_ERR("Total Stripe length(0x%llx)"
- " >= 32bit is not supported\n", _LLU(stripe_length));
- return -EINVAL;
- }
-
if (0 != (sbi->data_map.odm_stripe_unit & ~PAGE_MASK)) {
EXOFS_ERR("Stripe Unit(0x%llx)"
" must be Multples of PAGE_SIZE(0x%lx)\n",
@@ -360,8 +348,36 @@ static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
sbi->layout.stripe_unit = sbi->data_map.odm_stripe_unit;
sbi->layout.mirrors_p1 = sbi->data_map.odm_mirror_cnt + 1;
- sbi->layout.group_width = sbi->data_map.odm_num_comps /
+
+ if (sbi->data_map.odm_group_width) {
+ sbi->layout.group_width = sbi->data_map.odm_group_width;
+ sbi->layout.group_depth = sbi->data_map.odm_group_depth;
+ if (!sbi->layout.group_depth) {
+ EXOFS_ERR("group_depth == 0 && group_width != 0\n");
+ return -EINVAL;
+ }
+ sbi->layout.group_count = sbi->data_map.odm_num_comps /
+ sbi->layout.mirrors_p1 /
+ sbi->data_map.odm_group_width;
+ } else {
+ if (sbi->data_map.odm_group_depth) {
+ printk(KERN_NOTICE "Warning: group_depth ignored "
+ "group_width == 0 && group_depth == %d\n",
+ sbi->data_map.odm_group_depth);
+ sbi->data_map.odm_group_depth = 0;
+ }
+ sbi->layout.group_width = sbi->data_map.odm_num_comps /
sbi->layout.mirrors_p1;
+ sbi->layout.group_depth = -1;
+ sbi->layout.group_count = 1;
+ }
+
+ stripe_length = (u64)sbi->layout.group_width * sbi->layout.stripe_unit;
+ if (stripe_length >= (1ULL << 32)) {
+ EXOFS_ERR("Total Stripe length(0x%llx)"
+ " >= 32bit is not supported\n", _LLU(stripe_length));
+ return -EINVAL;
+ }
return 0;
}
@@ -540,6 +556,8 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
sbi->layout.stripe_unit = PAGE_SIZE;
sbi->layout.mirrors_p1 = 1;
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;