[Devel] [PATCH 10/10] ploop: Populate and maintain holes bitmap
Holes bitmap is needed for allocation of next free cluster.
Otherwise we don't know, where cluster number should be taken.
TODO: Flag to handle broken allocs (set bit back)
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/fmt_ploop1.c | 199 +-
drivers/block/ploop/map.c|6 +
include/linux/ploop/ploop.h |3 +
3 files changed, 205 insertions(+), 3 deletions(-)
diff --git a/drivers/block/ploop/fmt_ploop1.c b/drivers/block/ploop/fmt_ploop1.c
index 9cfc56296e5e..6212634a1c0e 100644
--- a/drivers/block/ploop/fmt_ploop1.c
+++ b/drivers/block/ploop/fmt_ploop1.c
@@ -23,6 +23,7 @@ struct ploop1_private
u64 bd_size;
u32 alloc_head;
sector_tl1_off;
+ u32 nr_bat_entries;
};
static unsigned int data_off_in_clusters(struct ploop_delta *delta)
@@ -78,6 +79,11 @@ static int ploop1_stop(struct ploop_delta * delta)
struct ploop_pvd_header *vh;
struct ploop1_private * ph = delta->priv;
+ if (delta->holes_bitmap) {
+ kvfree(delta->holes_bitmap);
+ delta->holes_bitmap = NULL;
+ }
+
if ((delta->flags & PLOOP_FMT_RDONLY) ||
test_bit(PLOOP_S_ABORT, >plo->state))
return 0;
@@ -115,6 +121,70 @@ ploop1_compose(struct ploop_delta * delta, int nchunks,
struct ploop_ctl_chunk *
return ploop_io_init(delta, nchunks, pc);
}
+static int populate_holes_bitmap(struct ploop_delta *delta,
+struct ploop1_private *ph)
+{
+ unsigned int block, nr_blocks, size, off;
+ struct page *page;
+ sector_t sec;
+ u32 *index;
+ int i, ret;
+
+ if (test_bit(PLOOP_S_NO_FALLOC_DISCARD, >plo->state))
+ return 0;
+
+ /* To do: add discard alignment for v1 */
+ if (delta->plo->fmt_version != PLOOP_FMT_V2)
+ return 0;
+
+ ret = -ENOMEM;
+ page = alloc_page(GFP_KERNEL);
+ if (!page)
+ return ret;
+
+ nr_blocks = ph->nr_bat_entries;
+ /* Bits to bytes */
+ size = DIV_ROUND_UP(nr_blocks, 8);
+
+ delta->holes_bitmap = kvmalloc(size, GFP_KERNEL);
+ if (!delta->holes_bitmap)
+ goto put_page;
+ memset(delta->holes_bitmap, 0xff, size);
+ for (i = nr_blocks; i < size * 8; i++)
+ clear_bit(i, delta->holes_bitmap);
+
+ block = 0;
+ while (block < nr_blocks) {
+ if (!ploop1_map_index(delta, block, )) {
+ WARN_ONCE(1, "Can't map block\n");
+ goto put_page;
+ }
+ ret = delta->io.ops->sync_read(>io, page,
+ 4096, 0, sec);
+ if (ret)
+ goto put_page;
+
+ off = block ? 0 : PLOOP_MAP_OFFSET;
+
+ index = page_address(page);
+ for (i = off; i < INDEX_PER_PAGE && block + i - off <
nr_blocks; i++) {
+ if (index[i] != 0) {
+ clear_bit((index[i] >>
ploop_map_log(delta->plo)) -
+ data_off_in_clusters(delta),
+ delta->holes_bitmap);
+ }
+ }
+
+ block += (block ? INDEX_PER_PAGE : INDEX_PER_PAGE -
PLOOP_MAP_OFFSET);
+ }
+
+ ret = 0;
+
+put_page:
+ put_page(page);
+ return ret;
+}
+
static int
ploop1_open(struct ploop_delta * delta)
{
@@ -157,6 +227,7 @@ ploop1_open(struct ploop_delta * delta)
goto out_err;
ph->l1_off = le32_to_cpu(vh->m_FirstBlockOffset);
+ ph->nr_bat_entries = le32_to_cpu(vh->m_Size);
err = -EBUSY;
if (pvd_header_is_disk_in_use(vh))
@@ -237,12 +308,37 @@ static void
ploop1_allocate(struct ploop_delta * delta, struct ploop_request * preq,
struct bio_list * sbl, unsigned int size)
{
- if (delta->io.alloc_head >=
+ struct ploop1_private * ph = delta->priv;
+ cluster_t cluster = 0;
+ int ret;
+
+ if (delta->holes_bitmap) {
+ unsigned nr_clusters = ph->nr_bat_entries;
+
+ cluster = find_first_bit(delta->holes_bitmap, nr_clusters);
+ if (cluster >= nr_clusters) {
+ PLOOP_FAIL_REQUEST(preq, -ENOSPC);
+ return;
+ }
+ cluster += data_off_in_clusters(delta);
+ } else if (delta->io.alloc_head >=
(delta->max_delta_size >> delta->cluster_log)) {
PLOOP_FAIL_REQUEST(preq, -E2BIG);
return;
}
- ploop_submit_alloc(delta, preq, sbl, size, 0);
+
+ ret = ploop_submit_alloc(delta, preq, sbl, size, cluster);
+
+ if (ret == 1 && cluster) {
+ /* Success. Mark cluster as occupied */
+ cluster -=
[Devel] [PATCH 09/10] ploop: Fallocate cluster in cached_submit() during hole reuse
__map_extent_bmap() is for raw format, when we don't have
information about presence of a cluster.
Ploop1 must allocate all the space in beginning of cached_submit()
function. Otherwise, we can't control what is going on.
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/io_direct.c | 27 +++
drivers/block/ploop/io_direct_map.c | 11 +++
2 files changed, 34 insertions(+), 4 deletions(-)
diff --git a/drivers/block/ploop/io_direct.c b/drivers/block/ploop/io_direct.c
index ad95cce91f6e..1667989c9c3b 100644
--- a/drivers/block/ploop/io_direct.c
+++ b/drivers/block/ploop/io_direct.c
@@ -389,21 +389,29 @@ cached_submit(struct ploop_io *io, iblock_t iblk, struct
ploop_request * preq,
loff_t pos, end_pos, start, end;
loff_t clu_siz = cluster_size_in_bytes(plo);
struct bio_iter biter;
- loff_t new_size;
+ loff_t new_size, prealloc;
loff_t used_pos;
bool may_fallocate = dio_may_fallocate(io);
+ bool reusing, once = true;
trace_cached_submit(preq);
pos = (loff_t)iblk << (plo->cluster_log + 9);
end_pos = pos + clu_siz;
used_pos = (loff_t)(io->alloc_head - 1) << (io->plo->cluster_log + 9);
+ reusing = (end_pos <= used_pos);
+
+ if (reusing) {
+ /* Reusing a hole */
+ prealloc = clu_siz;
+ goto try_again;
+ }
file_start_write(io->files.file);
- if (use_prealloc && end_pos > used_pos && may_fallocate) {
+ if (use_prealloc && (end_pos > used_pos) && may_fallocate) {
if (unlikely(io->prealloced_size < used_pos + clu_siz)) {
- loff_t prealloc = end_pos;
+ prealloc = end_pos;
if (prealloc > PLOOP_MAX_PREALLOC(plo))
prealloc = PLOOP_MAX_PREALLOC(plo);
try_again:
@@ -423,7 +431,8 @@ cached_submit(struct ploop_io *io, iblock_t iblk, struct
ploop_request * preq,
if (err)
goto end_write;
- io->prealloced_size = pos + prealloc;
+ if (io->prealloced_size < pos + prealloc)
+ io->prealloced_size = pos + prealloc;
}
}
@@ -434,6 +443,16 @@ cached_submit(struct ploop_io *io, iblock_t iblk, struct
ploop_request * preq,
if (unlikely(IS_ERR(em))) {
err = PTR_ERR(em);
+ if (err == -ENOENT && once) {
+ /*
+* Boundary cluster: temporary crutch
+* before io->alloc_head is reworked
+* to not be incremented in caller.
+*/
+ once = false;
+ prealloc = clu_siz;
+ goto try_again;
+ }
goto end_write;
}
diff --git a/drivers/block/ploop/io_direct_map.c
b/drivers/block/ploop/io_direct_map.c
index 7934bc8ecfb1..84a3986b8769 100644
--- a/drivers/block/ploop/io_direct_map.c
+++ b/drivers/block/ploop/io_direct_map.c
@@ -703,7 +703,18 @@ static struct extent_map *__map_extent_bmap(struct
ploop_io *io,
}
if (fieinfo.fi_extents_mapped != 1) {
+ struct ploop_device *plo = io->plo;
ploop_extent_put(em);
+ /*
+* In case of io_direct we may support discards
+* in multi-delta case, since all allocated blocks
+* are added to extent tree. But we follow generic
+* way, and encode discarded blocks by zeroing
+* their indexes in maps (ploop1).
+*/
+ if (!test_bit(PLOOP_MAP_IDENTICAL, >map.flags))
+ return ERR_PTR(-ENOENT);
+
ret = fallocate_cluster(io, inode, start_off, len,
align_to_clu);
if (!ret)
goto again;
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 07/10] ploop: Add .complete_merge method
It will be used to reallocated holes bitmap
after merge.
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/dev.c | 17 -
drivers/block/ploop/io_direct_map.c |2 ++
include/linux/ploop/ploop.h |1 +
3 files changed, 15 insertions(+), 5 deletions(-)
diff --git a/drivers/block/ploop/dev.c b/drivers/block/ploop/dev.c
index 57920a93129a..618d22bf02ec 100644
--- a/drivers/block/ploop/dev.c
+++ b/drivers/block/ploop/dev.c
@@ -3762,18 +3762,25 @@ static void ploop_update_fmt_version(struct
ploop_device * plo)
}
}
-static void ploop_merge_cleanup(struct ploop_device * plo,
- struct ploop_map * map,
- struct ploop_delta * delta, int err)
+static void ploop_merge_complete(struct ploop_device * plo,
+struct ploop_map * map,
+struct ploop_delta * delta, int err)
{
+ struct ploop_delta *top_delta;
+
ploop_quiesce(plo);
mutex_lock(>sysfs_mutex);
list_del(>list);
if (err)
list_add(>list, >map.delta_list);
- else
+ else {
+ top_delta = ploop_top_delta(plo);
+ if (top_delta->ops->complete_merge)
+ /* FIXME: How should we handle error here? */
+ top_delta->ops->complete_merge(top_delta);
ploop_update_fmt_version(plo);
+ }
plo->trans_map = NULL;
plo->maintenance_type = PLOOP_MNTN_OFF;
@@ -3873,7 +3880,7 @@ static int ploop_merge(struct ploop_device * plo)
err = -EIO;
}
- ploop_merge_cleanup(plo, map, delta, err);
+ ploop_merge_complete(plo, map, delta, err);
if (!err) {
kobject_del(>kobj);
diff --git a/drivers/block/ploop/io_direct_map.c
b/drivers/block/ploop/io_direct_map.c
index bc65e60e72a3..7934bc8ecfb1 100644
--- a/drivers/block/ploop/io_direct_map.c
+++ b/drivers/block/ploop/io_direct_map.c
@@ -620,6 +620,8 @@ static int fallocate_cluster(struct ploop_io *io, struct
inode *inode,
}
ret = file->f_op->fallocate(file, FALLOC_FL_KEEP_SIZE, start_off, len);
+ if (!ret)
+ ret = io->ops->sync(io);
out:
return ret;
}
diff --git a/include/linux/ploop/ploop.h b/include/linux/ploop/ploop.h
index 73141c536f65..a730c2de7873 100644
--- a/include/linux/ploop/ploop.h
+++ b/include/linux/ploop/ploop.h
@@ -283,6 +283,7 @@ struct ploop_delta_ops
int (*complete_snapshot)(struct ploop_delta *, struct
ploop_snapdata *);
int (*prepare_merge)(struct ploop_delta *, struct
ploop_snapdata *);
int (*start_merge)(struct ploop_delta *, struct
ploop_snapdata *);
+ int (*complete_merge)(struct ploop_delta *);
int (*truncate)(struct ploop_delta *, struct file *, __u32
alloc_head);
int (*prepare_grow)(struct ploop_delta *, u64 *new_size,
int *reloc);
int (*complete_grow)(struct ploop_delta *, u64 new_size);
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 08/10] ploop: Zero indexes on discard
Shitch preq into PLOOP_E_DATA_WBI state
to continue execution after discard's write
is finished.
Zero index in that stage.
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/dev.c|5 -
drivers/block/ploop/fmt_ploop1.c |5 +
2 files changed, 9 insertions(+), 1 deletion(-)
diff --git a/drivers/block/ploop/dev.c b/drivers/block/ploop/dev.c
index 618d22bf02ec..3ada41f91867 100644
--- a/drivers/block/ploop/dev.c
+++ b/drivers/block/ploop/dev.c
@@ -2461,7 +2461,10 @@ ploop_entry_request(struct ploop_request * preq)
spin_unlock_irq(>lock);
}
preq->iblock = iblk;
- preq->eng_state = PLOOP_E_COMPLETE;
+ if (!(preq->req_rw & REQ_DISCARD))
+ preq->eng_state = PLOOP_E_COMPLETE;
+ else
+ preq->eng_state = PLOOP_E_DATA_WBI;
__TRACE("T %p %u\n", preq, preq->req_cluster);
plo->st.bio_out++;
delta->io.ops->submit(>io, preq,
preq->req_rw,
diff --git a/drivers/block/ploop/fmt_ploop1.c b/drivers/block/ploop/fmt_ploop1.c
index 370aa6369a8a..9cfc56296e5e 100644
--- a/drivers/block/ploop/fmt_ploop1.c
+++ b/drivers/block/ploop/fmt_ploop1.c
@@ -250,6 +250,11 @@ ploop1_allocate(struct ploop_delta * delta, struct
ploop_request * preq,
static void
ploop1_allocate_complete(struct ploop_delta * delta, struct ploop_request *
preq)
{
+ if ((preq->req_rw & REQ_DISCARD) &&
+ !test_bit(PLOOP_REQ_DISCARD, >state)) {
+ preq->iblock = 0;
+ }
+
ploop_index_update(preq);
}
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 02/10] ploop: Make submit_alloc() return int value
-1 is error;
0 means this service preq is skipped;
1 is block was allocated
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/io_direct.c |6 --
drivers/block/ploop/io_kaio.c | 11 ++-
include/linux/ploop/ploop.h |2 +-
3 files changed, 11 insertions(+), 8 deletions(-)
diff --git a/drivers/block/ploop/io_direct.c b/drivers/block/ploop/io_direct.c
index 21cfdeb7441b..b1955d339de5 100644
--- a/drivers/block/ploop/io_direct.c
+++ b/drivers/block/ploop/io_direct.c
@@ -730,7 +730,7 @@ dio_submit_pad(struct ploop_io *io, struct ploop_request *
preq,
PLOOP_FAIL_REQUEST(preq, err);
}
-static void
+static int
dio_submit_alloc(struct ploop_io *io, struct ploop_request * preq,
struct bio_list * sbl, unsigned int size)
{
@@ -741,7 +741,7 @@ dio_submit_alloc(struct ploop_io *io, struct ploop_request
* preq,
if (!(io->files.file->f_mode & FMODE_WRITE)) {
PLOOP_FAIL_REQUEST(preq, -EBADF);
- return;
+ return -1;
}
err = cached_submit(io, iblk, preq, sbl, size, true);
@@ -749,8 +749,10 @@ dio_submit_alloc(struct ploop_io *io, struct ploop_request
* preq,
if (err == -ENOSPC)
io->alloc_head--;
PLOOP_FAIL_REQUEST(preq, err);
+ return -1;
}
preq->eng_state = PLOOP_E_DATA_WBI;
+ return 1;
}
/* When backing fs does not export any method to allocate new blocks
diff --git a/drivers/block/ploop/io_kaio.c b/drivers/block/ploop/io_kaio.c
index 84da1cec5643..1e4ae6efbe70 100644
--- a/drivers/block/ploop/io_kaio.c
+++ b/drivers/block/ploop/io_kaio.c
@@ -546,7 +546,7 @@ static int kaio_fsync_thread(void * data)
return 0;
}
-static void
+static int
kaio_submit_alloc(struct ploop_io *io, struct ploop_request * preq,
struct bio_list * sbl, unsigned int size)
{
@@ -558,7 +558,7 @@ kaio_submit_alloc(struct ploop_io *io, struct ploop_request
* preq,
if (delta->flags & PLOOP_FMT_RDONLY) {
PLOOP_FAIL_REQUEST(preq, -EBADF);
- return;
+ return -1;
}
iblk = io->alloc_head;
@@ -568,7 +568,7 @@ kaio_submit_alloc(struct ploop_io *io, struct ploop_request
* preq,
kaio_queue_fsync_req(preq);
io->plo->st.bio_syncwait++;
spin_unlock_irq(>plo->lock);
- return;
+ return 0;
}
BUG_ON(preq->prealloc_size);
@@ -586,11 +586,11 @@ kaio_submit_alloc(struct ploop_io *io, struct
ploop_request * preq,
kaio_queue_trunc_req(preq);
io->plo->st.bio_syncwait++;
spin_unlock_irq(>plo->lock);
- return;
+ return 0;
} else { /* we're not first */
list_add_tail(>list,
>prealloc_preq->delay_list);
- return;
+ return 0;
}
}
@@ -600,6 +600,7 @@ kaio_submit_alloc(struct ploop_io *io, struct ploop_request
* preq,
preq->eng_state = PLOOP_E_DATA_WBI;
kaio_sbl_submit(io->files.file, preq, REQ_WRITE, sbl, iblk, size);
+ return 1;
}
static int kaio_release_prealloced(struct ploop_io * io)
diff --git a/include/linux/ploop/ploop.h b/include/linux/ploop/ploop.h
index b209c9f096a8..599f7dbff7f2 100644
--- a/include/linux/ploop/ploop.h
+++ b/include/linux/ploop/ploop.h
@@ -159,7 +159,7 @@ struct ploop_io_ops
void(*submit)(struct ploop_io *, struct ploop_request *,
unsigned long rw,
struct bio_list *sbl, iblock_t iblk, unsigned int
size);
- void(*submit_alloc)(struct ploop_io *, struct ploop_request *,
+ int (*submit_alloc)(struct ploop_io *, struct ploop_request *,
struct bio_list *sbl, unsigned int size);
void(*post_submit)(struct ploop_io *, struct ploop_request *);
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 01/10] ploop: Export map defines to separate header file
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/fmt_ploop1.c |4 +--
drivers/block/ploop/map.c| 52 +--
drivers/block/ploop/map.h| 56 ++
3 files changed, 58 insertions(+), 54 deletions(-)
create mode 100644 drivers/block/ploop/map.h
diff --git a/drivers/block/ploop/fmt_ploop1.c b/drivers/block/ploop/fmt_ploop1.c
index 1281a343bff4..382737b4cb6c 100644
--- a/drivers/block/ploop/fmt_ploop1.c
+++ b/drivers/block/ploop/fmt_ploop1.c
@@ -12,13 +12,11 @@
#include
#include "ploop1_image.h"
+#include "map.h"
/* The implementaion of ploop1 (PVD) delta format, defined in ploop1_fmt.h
*/
-#define INDEX_PER_PAGE (PAGE_SIZE / 4)
-#define INDEX_PER_PAGE_SHIFT (PAGE_SHIFT - 2)
-
struct ploop1_private
{
struct page *dyn_page;
diff --git a/drivers/block/ploop/map.c b/drivers/block/ploop/map.c
index 7b08001aa58b..8a28148c17cb 100644
--- a/drivers/block/ploop/map.c
+++ b/drivers/block/ploop/map.c
@@ -40,12 +40,7 @@
#include
#include
-
-/* This defines slot in mapping page. Right now it is 32 bit
- * and therefore it directly matches ploop1 structure. */
-typedef u32 map_index_t;
-
-#define INDEX_PER_PAGE (PAGE_SIZE / sizeof(map_index_t))
+#include "map.h"
static struct kmem_cache * ploop_map_cache;
@@ -53,51 +48,6 @@ static LIST_HEAD(map_lru);
static DEFINE_SPINLOCK(map_lru_lock);
static atomic_t map_pages_nr = ATOMIC_INIT(0);
-/*
- * Additional information for each page is:
- * 1. rb tree link
- * 2. Page
- * 3. mn_start, mn_end - the first and the last index
- * (correspondingly) the page maps to iblocks.
- * 4. lru linkage
- * 5. delta level of whole page, it is delta, where this page
- *is backed.
- * 6. Array of delta levels for each map_index in the page.
- *If page is backed at level N, those levels cannot be >N.
- *If all the levels == N, array of levels is not allocated.
- *When at least one level < N, it is stored in the array.
- *Note, that in this case exporting page to disk implies
- *clearing irrelevant entries.
- */
-
-struct map_node
-{
- struct rb_node rb_link;
- cluster_t mn_start;
- cluster_t mn_end;
- unsigned long state;
- atomic_trefcnt;
- struct ploop_map*parent;
-
- struct page *page;
- struct list_headlru;
- u8 *levels;
-
- /* List of preq's blocking on this mapping.
-*
-* We queue here several kinds of requests:
-* 1. If mapping is not uptodate, all the requests which need
-*this mapping are queued here. preq state is ENTRY.
-* 2. If preq requires index update and it is delayed
-*because writeback is in progress. preq state is INDEX_DELAY,
-*new index is kept in preq->iblock.
-* 3. If preq's started index update, preq state is INDEX_WB,
-*new indices are sent to io, but they are not inserted
-*into mapping until writeback is complete.
-*/
- struct list_headio_queue;
-};
-
cluster_t map_get_mn_end(struct map_node *m)
{
return m->mn_end;
diff --git a/drivers/block/ploop/map.h b/drivers/block/ploop/map.h
new file mode 100644
index ..580f23ccfbb0
--- /dev/null
+++ b/drivers/block/ploop/map.h
@@ -0,0 +1,56 @@
+#ifndef __PLOOP_MAP_H
+#define __PLOOP_MAP_H
+
+/* This defines slot in mapping page. Right now it is 32 bit
+ * and therefore it directly matches ploop1 structure. */
+typedef u32 map_index_t;
+
+#define INDEX_PER_PAGE (PAGE_SIZE / sizeof(map_index_t))
+#define INDEX_PER_PAGE_SHIFT (PAGE_SHIFT - 2)
+
+/*
+ * Additional information for each page is:
+ * 1. rb tree link
+ * 2. Page
+ * 3. mn_start, mn_end - the first and the last index
+ * (correspondingly) the page maps to iblocks.
+ * 4. lru linkage
+ * 5. delta level of whole page, it is delta, where this page
+ *is backed.
+ * 6. Array of delta levels for each map_index in the page.
+ *If page is backed at level N, those levels cannot be >N.
+ *If all the levels == N, array of levels is not allocated.
+ *When at least one level < N, it is stored in the array.
+ *Note, that in this case exporting page to disk implies
+ *clearing irrelevant entries.
+ */
+
+struct map_node
+{
+ struct rb_node rb_link;
+ cluster_t mn_start;
+ cluster_t mn_end;
+ unsigned long state;
+ atomic_trefcnt;
+ struct ploop_map*parent;
+
+ struct page *page;
+ struct list_headlru;
+ u8 *levels;
+
+ /* List of preq's blocking on this mapping.
+*
+* We queue here several kinds of requests:
+* 1. If mapping is not uptodate, all the requests which need
+*this mapping are
[Devel] [PATCH 04/10] ploop: Prohibit discard ioctls
Prohibit maintaince mode related ioctls.
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/dev.c | 10 ++
drivers/block/ploop/discard.c |6 ++
2 files changed, 16 insertions(+)
diff --git a/drivers/block/ploop/dev.c b/drivers/block/ploop/dev.c
index f8b9b0bcba78..11bc864a5679 100644
--- a/drivers/block/ploop/dev.c
+++ b/drivers/block/ploop/dev.c
@@ -4522,6 +4522,8 @@ static int ploop_freeblks_ioc(struct ploop_device *plo,
unsigned long arg)
int i;
int rc = 0;
+ return -EINVAL;
+
if (list_empty(>map.delta_list))
return -ENOENT;
@@ -4605,6 +4607,8 @@ static int ploop_fbget_ioc(struct ploop_device *plo,
unsigned long arg)
struct ploop_freeblks_ctl ctl;
int rc = 0;
+ return -EINVAL;
+
if (list_empty(>map.delta_list))
return -ENOENT;
@@ -4629,6 +4633,8 @@ static int ploop_fbfilter_ioc(struct ploop_device *plo,
unsigned long arg)
{
int rc = 0;
+ return -EINVAL;
+
if (plo->maintenance_type != PLOOP_MNTN_DISCARD ||
!test_bit(PLOOP_S_DISCARD_LOADED, >state))
return -EINVAL;
@@ -4716,6 +4722,8 @@ static void ploop_discard_restart(struct ploop_device
*plo, int err)
static int ploop_fbdrop_ioc(struct ploop_device *plo)
{
+ return -EINVAL;
+
if (list_empty(>map.delta_list))
return -ENOENT;
@@ -4742,6 +4750,8 @@ static int ploop_relocblks_ioc(struct ploop_device *plo,
unsigned long arg)
int err = 0;
int n_free;
+ return -EINVAL;
+
if (list_empty(>map.delta_list))
return -ENOENT;
diff --git a/drivers/block/ploop/discard.c b/drivers/block/ploop/discard.c
index 33122497b9c5..1920ab05e20b 100644
--- a/drivers/block/ploop/discard.c
+++ b/drivers/block/ploop/discard.c
@@ -17,6 +17,8 @@ int ploop_discard_init_ioc(struct ploop_device *plo)
struct ploop_freeblks_desc *fbd;
struct ploop_delta *delta = ploop_top_delta(plo);
+ return -EINVAL;
+
if (delta == NULL)
return -EINVAL;
@@ -52,6 +54,8 @@ int ploop_discard_fini_ioc(struct ploop_device *plo)
struct ploop_request *preq, *tmp;
LIST_HEAD(drop_list);
+ return -EINVAL;
+
if (!test_and_clear_bit(PLOOP_S_DISCARD, >state))
return 0;
@@ -90,6 +94,8 @@ int ploop_discard_wait_ioc(struct ploop_device *plo)
{
int err;
+ return -EINVAL;
+
if (!test_bit(PLOOP_S_DISCARD, >state))
return 0;
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 06/10] ploop: Introduce data_off_in_clusters() helper
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/fmt_ploop1.c |7 +++
1 file changed, 7 insertions(+)
diff --git a/drivers/block/ploop/fmt_ploop1.c b/drivers/block/ploop/fmt_ploop1.c
index eea15aac8790..370aa6369a8a 100644
--- a/drivers/block/ploop/fmt_ploop1.c
+++ b/drivers/block/ploop/fmt_ploop1.c
@@ -25,6 +25,13 @@ struct ploop1_private
sector_tl1_off;
};
+static unsigned int data_off_in_clusters(struct ploop_delta *delta)
+{
+ struct ploop1_private *ph = delta->priv;
+
+ return (ph->l1_off >> delta->cluster_log);
+}
+
int ploop1_map_index(struct ploop_delta * delta, unsigned long block, sector_t
*sec)
{
struct ploop1_private * ph = delta->priv;
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 00/10] ploop: Discard with zeroing of ploop1 indexes support
--- Kirill Tkhai (10): ploop: Export map defines to separate header file ploop: Make submit_alloc() return int value ploop: Introduce ploop_submit_alloc() helper ploop: Prohibit discard ioctls ploop: Prohibit PLOOP_IOC_UPDATE_INDEX on singular list ploop: Introduce data_off_in_clusters() helper ploop: Add .complete_merge method ploop: Zero indexes on discard ploop: Fallocate cluster in cached_submit() during hole reuse ploop: Populate and maintain holes bitmap drivers/block/ploop/dev.c | 41 ++- drivers/block/ploop/discard.c |6 + drivers/block/ploop/fmt_ploop1.c| 215 ++- drivers/block/ploop/fmt_raw.c |2 drivers/block/ploop/io.c| 39 ++ drivers/block/ploop/io_direct.c | 43 --- drivers/block/ploop/io_direct_map.c | 13 ++ drivers/block/ploop/io_kaio.c | 22 +--- drivers/block/ploop/map.c | 58 + drivers/block/ploop/map.h | 56 + include/linux/ploop/ploop.h | 11 +- 11 files changed, 406 insertions(+), 100 deletions(-) create mode 100644 drivers/block/ploop/map.h -- Signed-off-by: Kirill Tkhai ___ Devel mailing list Devel@openvz.org https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 05/10] ploop: Prohibit PLOOP_IOC_UPDATE_INDEX on singular list
Add debug check to not miss a moment userspace decides to change its behavior. Signed-off-by: Kirill Tkhai --- drivers/block/ploop/dev.c |9 - 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/block/ploop/dev.c b/drivers/block/ploop/dev.c index 11bc864a5679..57920a93129a 100644 --- a/drivers/block/ploop/dev.c +++ b/drivers/block/ploop/dev.c @@ -4277,7 +4277,14 @@ static int ploop_index_update_ioc(struct ploop_device *plo, unsigned long arg) struct reloc_map *map; int i; - if (list_empty(>map.delta_list)) + /* +* Currently this is used by userspace for relocation +* of start blocks before expanding BAT of !top delta. +* In case of someone wants to do that on a singular +* list, he should expand holes bitmap too. +*/ + if (list_empty(>map.delta_list) || + list_is_singular(>map.delta_list)) return -ENOENT; if (copy_from_user(, (void*)arg, ___ Devel mailing list Devel@openvz.org https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 0/3] fix jumbo chunk warning
Initially the warning was added incorrectly due to unsynchronization of pcs_mds_sys_info structure with the userspace client. Let's sync the structure and fix that. Pavel Butsykin (3): Revert "fs/fuse kio: add warning about jumbo chunks" fs/fuse kio: sync pcs_mds_sys_info struct fs/fuse kio: add warning about jumbo chunks fs/fuse/kio/pcs/pcs_client_types.h | 6 +++--- fs/fuse/kio/pcs/pcs_cluster.c | 8 fs/fuse/kio/pcs/pcs_fuse_kdirect.c | 4 ++-- fs/fuse/kio/pcs/pcs_mds_prot.h | 2 -- fs/fuse/kio/pcs/pcs_prot_types.h | 10 ++ 5 files changed, 15 insertions(+), 15 deletions(-) -- 2.15.1 ___ Devel mailing list Devel@openvz.org https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 2/3] fs/fuse kio: sync pcs_mds_sys_info struct
For some reason pcs_mds_sys_info structure is different in the kernel and
userspace. Let's synchronize it to avoid inaccuracies and discrepancies in
the future.
Signed-off-by: Pavel Butsykin
---
fs/fuse/kio/pcs/pcs_client_types.h | 6 +++---
fs/fuse/kio/pcs/pcs_cluster.c | 8
fs/fuse/kio/pcs/pcs_fuse_kdirect.c | 2 +-
fs/fuse/kio/pcs/pcs_prot_types.h | 10 ++
4 files changed, 14 insertions(+), 12 deletions(-)
diff --git a/fs/fuse/kio/pcs/pcs_client_types.h
b/fs/fuse/kio/pcs/pcs_client_types.h
index 1be32cbbf285..c0d14ec5b5d4 100644
--- a/fs/fuse/kio/pcs/pcs_client_types.h
+++ b/fs/fuse/kio/pcs/pcs_client_types.h
@@ -87,9 +87,9 @@ static inline void pcs_set_fileinfo(struct pcs_dentry_info
*i, const struct pcs_
if (mi->sys.stripe_depth == 0) {
mi->sys.stripe_depth = 1;
- mi->sys.strip_width = mi->sys.chunk_size;
+ mi->sys.strip_width = mi->sys.chunk_size_lo;
}
- i->mapping.chunk_size_bits = ilog2(mi->sys.chunk_size);
+ i->mapping.chunk_size_bits = ilog2(mi->sys.chunk_size_lo);
}
@@ -170,7 +170,7 @@ typedef struct _pcs_api_csconnreq_t {
#define DENTRY_ARGS(de) PCS_FILE_ID_ARGS((de)->id.parent),
PCS_FILE_ID_ARGS((de)->fileinfo.attr.id)
#define DENTRY_SIZE(de) ((de)->fileinfo.attr.size)
-#define DENTRY_CHUNK_SIZE(de) ((de)->fileinfo.sys.chunk_size)
+#define DENTRY_CHUNK_SIZE(de) ((de)->fileinfo.sys.chunk_size_lo)
#define DENTRY_CHUNK_SIZE_BITS(de) ((de)->mapping.chunk_size_bits)
void pcs_mapset_limit(struct pcs_map_set *maps, int limit);
diff --git a/fs/fuse/kio/pcs/pcs_cluster.c b/fs/fuse/kio/pcs/pcs_cluster.c
index 5df263f01f98..a73120c97e5e 100644
--- a/fs/fuse/kio/pcs/pcs_cluster.c
+++ b/fs/fuse/kio/pcs/pcs_cluster.c
@@ -364,9 +364,9 @@ static noinline void __pcs_cc_process_ireq_rw(struct
pcs_int_request *ireq)
unsigned int len;
u64 rpos, chunk, end_pos;
- rpos = map_file_to_chunk(pos, di->fileinfo.sys.chunk_size,
di->fileinfo.sys.stripe_depth, di->fileinfo.sys.strip_width);
+ rpos = map_file_to_chunk(pos, di->fileinfo.sys.chunk_size_lo,
di->fileinfo.sys.stripe_depth, di->fileinfo.sys.strip_width);
- chunk = rpos & ~((u64)di->fileinfo.sys.chunk_size - 1);
+ chunk = rpos & ~((u64)di->fileinfo.sys.chunk_size_lo - 1);
end_pos = ((rpos / di->fileinfo.sys.strip_width) + 1) *
(u64)di->fileinfo.sys.strip_width;
sreq = ireq_alloc(di);
@@ -385,9 +385,9 @@ static noinline void __pcs_cc_process_ireq_rw(struct
pcs_int_request *ireq)
sreq->iochunk.cmd = ireq->apireq.req->type;
sreq->iochunk.cs_index = 0;
sreq->iochunk.chunk = chunk;
- sreq->iochunk.offset = rpos % di->fileinfo.sys.chunk_size;
+ sreq->iochunk.offset = rpos % di->fileinfo.sys.chunk_size_lo;
sreq->iochunk.dio_offset = dio_offset;
- len = di->fileinfo.sys.chunk_size - sreq->iochunk.offset;
+ len = di->fileinfo.sys.chunk_size_lo - sreq->iochunk.offset;
if (len > sz)
len = sz;
if (rpos + len > end_pos)
diff --git a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
index 875ad18df4a7..6d12b8038e8a 100644
--- a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
+++ b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
@@ -343,7 +343,7 @@ static int kpcs_do_file_open(struct fuse_conn *fc, struct
file *file, struct ino
INIT_LIST_HEAD(>kq);
spin_lock_init(>kq_lock);
TRACE("init id:%llu chunk_size:%d stripe_depth:%d strip_width:%d\n",
- fi->nodeid, di->fileinfo.sys.chunk_size,
+ fi->nodeid, di->fileinfo.sys.chunk_size_lo,
di->fileinfo.sys.stripe_depth, di->fileinfo.sys.strip_width);
ret = fuse_pcs_kdirect_claim_op(fc, file, true);
diff --git a/fs/fuse/kio/pcs/pcs_prot_types.h b/fs/fuse/kio/pcs/pcs_prot_types.h
index bba1f1dc76e3..0e7bc9743885 100644
--- a/fs/fuse/kio/pcs/pcs_prot_types.h
+++ b/fs/fuse/kio/pcs/pcs_prot_types.h
@@ -104,15 +104,17 @@ struct __pre_aligned(8) pcs_mds_fattr
};
struct __pre_aligned(8) pcs_mds_sys_info {
- u32 map_type; /* reserved for RAID */
- u32 chunk_size; /* global constant */
+ u8 map_type; /* reserved for RAID */
+ u8 reserved[2];
+ u8 chunk_size_hi; /* chunk size (hi bits) */
+ u32 chunk_size_lo; /* chunk size (lo bits) */
u8 stripe_depth; /* for RAID6/RS */
u8 redundancy; /* number of checksums for RAID6/RS */
u8 tolerance;/* write-tolerance (how much lost replicas we can
tolerate still allowing writing) */
- u8 reserved8;
+ u8 reserved2;
u32 strip_width; /* length of strip for RAID6/RS */
u32 lease_tout; /* lease expiration timeout (in milliseconds) */
-
[Devel] [PATCH 1/3] Revert "fs/fuse kio: add warning about jumbo chunks"
This reverts commit afb8d534110dbe203e4ae6385cef79b38c9e4771.
Signed-off-by: Pavel Butsykin
---
fs/fuse/kio/pcs/pcs_fuse_kdirect.c | 3 ---
fs/fuse/kio/pcs/pcs_mds_prot.h | 2 --
2 files changed, 5 deletions(-)
diff --git a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
index 104fab25692b..875ad18df4a7 100644
--- a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
+++ b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
@@ -318,9 +318,6 @@ static int kpcs_do_file_open(struct fuse_conn *fc, struct
file *file, struct ino
if (info.sys.map_type != PCS_MAP_PLAIN) {
TRACE("Unsupported map_type:%x, ignore\n", info.sys.map_type);
-
- if (info.sys.map_type & PCS_JUMBO_CHUNK_FLAG)
- pr_warn_once("kio: fpath doesn't support jumbo
chunks\n");
return 0;
}
diff --git a/fs/fuse/kio/pcs/pcs_mds_prot.h b/fs/fuse/kio/pcs/pcs_mds_prot.h
index 4aca1e970170..80c20fde1537 100644
--- a/fs/fuse/kio/pcs/pcs_mds_prot.h
+++ b/fs/fuse/kio/pcs/pcs_mds_prot.h
@@ -259,8 +259,6 @@ enum
PCS_MAP_LS = PCS_MAP_COMBINED, /* Log structured storage */
};
-#define PCS_JUMBO_CHUNK_FLAG (1ULL << 63) /* Chunks size > 4G */
-
/* Max inline file size */
#define PCS_MAX_INLINE_SIZE 0x10 /* 1Mb */
--
2.15.1
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH 3/3] fs/fuse kio: add warning about jumbo chunks
KIO doesn't support jumbo chunks yet, so all requests to jumbo chunks are
silently redirected to user-space. It will be useful to see a message about
this until support has been added to KIO.
Signed-off-by: Pavel Butsykin
---
fs/fuse/kio/pcs/pcs_fuse_kdirect.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
index 6d12b8038e8a..aa1c79f440cb 100644
--- a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
+++ b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
@@ -318,6 +318,9 @@ static int kpcs_do_file_open(struct fuse_conn *fc, struct
file *file, struct ino
if (info.sys.map_type != PCS_MAP_PLAIN) {
TRACE("Unsupported map_type:%x, ignore\n", info.sys.map_type);
+
+ if (info.sys.chunk_size_hi)
+ pr_warn_once("kio: fpath doesn't support jumbo
chunks\n");
return 0;
}
--
2.15.1
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH] fs/fuse kio: fix getting a non-existent pcs inode in kpcs_kill_requests()
Some files may not have pcs inode, so it will be safer to check for !fi->private
before using it.
Signed-off-by: Pavel Butsykin
---
fs/fuse/kio/pcs/pcs_fuse_kdirect.c | 7 ++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
index 875ad18df4a7..a8c235244506 100644
--- a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
+++ b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
@@ -1506,11 +1506,16 @@ static void kpcs_kill_requests(struct fuse_conn *fc,
struct inode *inode)
list_for_each_entry(ff, >conn_files, fl) {
struct pcs_dentry_info *di;
+ struct fuse_inode *fi;
if (!ff->ff_dentry)
continue;
- di = get_pcs_inode(ff->ff_dentry->d_inode);
+ fi = get_fuse_inode(ff->ff_dentry->d_inode);
+ if (!fi->private)
+ continue;
+
+ di = pcs_inode_from_fuse(fi);
spin_lock(>kq_lock);
fuse_kill_requests(fc, inode, >kq);
--
2.15.1
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH RH7] ve: don't hide cgroup mount root from pseudosuper
Will need it to check mouninfo persists at the late restore, where the
root task is inside ve.
https://jira.sw.ru/browse/PSBM-86686
Signed-off-by: Pavel Tikhomirov
---
kernel/cgroup.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index d35e2ac150a3..9ca8af9f73d9 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -1862,7 +1862,9 @@ EXPORT_SYMBOL_GPL(cgroup_path);
int cgroup_path_ve(const struct cgroup *cgrp, char *buf, int buflen)
{
- return __cgroup_path(cgrp, buf, buflen, !ve_is_super(get_exec_env()));
+ struct ve_struct *ve = get_exec_env();
+
+ return __cgroup_path(cgrp, buf, buflen, !ve_is_super(ve) &&
!ve->is_pseudosuper);
}
/*
--
2.20.1
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH RHEL7 COMMIT] fs/fuse kio: fix double free of pfc in process_pcs_init_reply()
The commit is pushed to "branch-rh7-3.10.0-957.10.1.vz7.85.x-ovz" and will
appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-957.10.1.vz7.85.3
-->
commit 57a6200cbb664031506df3b2828f34b469e8d333
Author: Pavel Butsykin
Date: Tue Mar 26 12:49:03 2019 +0300
fs/fuse kio: fix double free of pfc in process_pcs_init_reply()
Remove redundant kvfree(pfc) in process_pcs_init_reply() to fix it.
This branch can only be taken if an abortion occurs during the mount, that
is
quite a rare event.
https://pmc.acronis.com/browse/VSTOR-20922
Signed-off-by: Pavel Butsykin
Reviewed-by: Kirill Tkhai
---
fs/fuse/kio/pcs/pcs_fuse_kdirect.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
index b908337f0e11..035f58302be3 100644
--- a/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
+++ b/fs/fuse/kio/pcs/pcs_fuse_kdirect.c
@@ -127,7 +127,6 @@ static void process_pcs_init_reply(struct fuse_conn *fc,
struct fuse_req *req)
if (pfc) {
fuse_ktrace_remove(fc);
pcs_cluster_fini(pfc);
- kvfree(pfc);
}
out:
if (fc->conn_error)
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
[Devel] [PATCH RHEL7 COMMIT] Revert "ploop: Remove excess variable"
The commit is pushed to "branch-rh7-3.10.0-957.10.1.vz7.85.x-ovz" and will
appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-957.10.1.vz7.85.3
-->
commit 855b630cf3d5afe7c96c230b1958d0c02ffd6ae6
Author: Kirill Tkhai
Date: Tue Mar 26 12:36:59 2019 +0300
Revert "ploop: Remove excess variable"
saved_ioc may be NULL. Then we increment the counter
and never decrement it. It even may overflow later.
https://jira.sw.ru/browse/PSBM-93006
Signed-off-by: Kirill Tkhai
---
drivers/block/ploop/dev.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/block/ploop/dev.c b/drivers/block/ploop/dev.c
index 1d6ee2329bb9..f8b9b0bcba78 100644
--- a/drivers/block/ploop/dev.c
+++ b/drivers/block/ploop/dev.c
@@ -2565,6 +2565,7 @@ static void ploop_req_state_process(struct ploop_request
* preq)
struct ploop_device * plo = preq->plo;
struct ploop_delta * top_delta;
struct io_context * saved_ioc = NULL;
+ int release_ioc = 0;
#ifdef CONFIG_BEANCOUNTERS
struct user_beancounter *saved_ub = NULL;
#endif
@@ -2575,6 +2576,7 @@ static void ploop_req_state_process(struct ploop_request
* preq)
saved_ioc = current->io_context;
current->io_context = preq->ioc;
atomic_long_inc(>ioc->refcount);
+ release_ioc = 1;
}
#ifdef CONFIG_BEANCOUNTERS
WARN_ONCE(!preq->preq_ub,
@@ -2906,7 +2908,7 @@ static void ploop_req_state_process(struct ploop_request
* preq)
BUG();
}
out:
- if (saved_ioc) {
+ if (release_ioc) {
struct io_context * ioc = current->io_context;
current->io_context = saved_ioc;
put_io_context(ioc);
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
