kmap() is being deprecated in favor of kmap_local_page().
Two main problems with kmap(): (1) It comes with an overhead as mapping
space is restricted and protected by a global lock for synchronization and
(2) it also requires global TLB invalidation when the kmap’s pool wraps
and it might block when the mapping space is fully utilized until a slot
becomes available.
With kmap_local_page() the mappings are per thread, CPU local, can take
page faults, and can be called from any context (including interrupts).
It is faster than kmap() in kernels with HIGHMEM enabled. Furthermore,
the tasks can be preempted and, when they are scheduled to run again, the
kernel virtual addresses are restored and still valid.
Since its use in bnode.c is safe everywhere, it should be preferred.
Therefore, replace kmap() with kmap_local_page() in bnode.c. Where
possible, use the suited standard helpers (memzero_page(), memcpy_page())
instead of open coding kmap_local_page() plus memset() or memcpy().
Suggested-by: Ira Weiny <[email protected]>
Signed-off-by: Fabio M. De Francesco <[email protected]>
---
fs/hfsplus/bnode.c | 105 +++++++++++++++++++++------------------------
1 file changed, 48 insertions(+), 57 deletions(-)
diff --git a/fs/hfsplus/bnode.c b/fs/hfsplus/bnode.c
index 177fae4e6581..3a1c77d0df48 100644
--- a/fs/hfsplus/bnode.c
+++ b/fs/hfsplus/bnode.c
@@ -29,14 +29,12 @@ void hfs_bnode_read(struct hfs_bnode *node, void *buf, int off, int len)
off &= ~PAGE_MASK;
l = min_t(int, len, PAGE_SIZE - off);
- memcpy(buf, kmap(*pagep) + off, l);
- kunmap(*pagep);
+ memcpy_from_page(buf, *pagep, off, l);
while ((len -= l) != 0) {
buf += l;
l = min_t(int, len, PAGE_SIZE);
- memcpy(buf, kmap(*++pagep), l);
- kunmap(*pagep);
+ memcpy_from_page(buf, *++pagep, 0, l);
}
}
@@ -82,16 +80,14 @@ void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
off &= ~PAGE_MASK;
l = min_t(int, len, PAGE_SIZE - off);
- memcpy(kmap(*pagep) + off, buf, l);
+ memcpy_to_page(*pagep, off, buf, l);
set_page_dirty(*pagep);
- kunmap(*pagep);
while ((len -= l) != 0) {
buf += l;
l = min_t(int, len, PAGE_SIZE);
- memcpy(kmap(*++pagep), buf, l);
+ memcpy_to_page(*++pagep, 0, buf, l);
set_page_dirty(*pagep);
- kunmap(*pagep);
}
}
@@ -112,15 +108,13 @@ void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
off &= ~PAGE_MASK;
l = min_t(int, len, PAGE_SIZE - off);
- memset(kmap(*pagep) + off, 0, l);
+ memzero_page(*pagep, off, l);
set_page_dirty(*pagep);
- kunmap(*pagep);
while ((len -= l) != 0) {
l = min_t(int, len, PAGE_SIZE);
- memset(kmap(*++pagep), 0, l);
+ memzero_page(*++pagep, 0, l);
set_page_dirty(*pagep);
- kunmap(*pagep);
}
}
@@ -142,24 +136,20 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
if (src == dst) {
l = min_t(int, len, PAGE_SIZE - src);
- memcpy(kmap(*dst_page) + src, kmap(*src_page) + src, l);
- kunmap(*src_page);
+ memcpy_page(*dst_page, src, *src_page, src, l);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
while ((len -= l) != 0) {
l = min_t(int, len, PAGE_SIZE);
- memcpy(kmap(*++dst_page), kmap(*++src_page), l);
- kunmap(*src_page);
+ memcpy_page(*++dst_page, 0, *++src_page, 0, l);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
}
} else {
void *src_ptr, *dst_ptr;
do {
- src_ptr = kmap(*src_page) + src;
- dst_ptr = kmap(*dst_page) + dst;
+ dst_ptr = kmap_local_page(*dst_page) + dst;
+ src_ptr = kmap_local_page(*src_page) + src;
if (PAGE_SIZE - src < PAGE_SIZE - dst) {
l = PAGE_SIZE - src;
src = 0;
@@ -171,9 +161,9 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
}
l = min(len, l);
memcpy(dst_ptr, src_ptr, l);
- kunmap(*src_page);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
if (!dst)
dst_page++;
else
@@ -185,6 +175,7 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
{
struct page **src_page, **dst_page;
+ void *src_ptr, *dst_ptr;
int l;
hfs_dbg(BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len);
@@ -202,27 +193,28 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
if (src == dst) {
while (src < len) {
- memmove(kmap(*dst_page), kmap(*src_page), src);
- kunmap(*src_page);
+ dst_ptr = kmap_local_page(*dst_page);
+ src_ptr = kmap_local_page(*src_page);
+ memmove(dst_ptr, src_ptr, src);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
len -= src;
src = PAGE_SIZE;
src_page--;
dst_page--;
}
src -= len;
- memmove(kmap(*dst_page) + src,
- kmap(*src_page) + src, len);
- kunmap(*src_page);
+ dst_ptr = kmap_local_page(*dst_page);
+ src_ptr = kmap_local_page(*src_page);
+ memmove(dst_ptr + src, src_ptr + src, len);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
} else {
- void *src_ptr, *dst_ptr;
-
do {
- src_ptr = kmap(*src_page) + src;
- dst_ptr = kmap(*dst_page) + dst;
+ dst_ptr = kmap_local_page(*dst_page) + dst;
+ src_ptr = kmap_local_page(*src_page) + src;
if (src < dst) {
l = src;
src = PAGE_SIZE;
@@ -234,9 +226,9 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
}
l = min(len, l);
memmove(dst_ptr - l, src_ptr - l, l);
- kunmap(*src_page);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
if (dst == PAGE_SIZE)
dst_page--;
else
@@ -251,26 +243,27 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
if (src == dst) {
l = min_t(int, len, PAGE_SIZE - src);
- memmove(kmap(*dst_page) + src,
- kmap(*src_page) + src, l);
- kunmap(*src_page);
+
+ dst_ptr = kmap_local_page(*dst_page) + src;
+ src_ptr = kmap_local_page(*src_page) + src;
+ memmove(dst_ptr, src_ptr, l);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
while ((len -= l) != 0) {
l = min_t(int, len, PAGE_SIZE);
- memmove(kmap(*++dst_page),
- kmap(*++src_page), l);
- kunmap(*src_page);
+ dst_ptr = kmap_local_page(*++dst_page);
+ src_ptr = kmap_local_page(*++src_page);
+ memmove(dst_ptr, src_ptr, l);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
}
} else {
- void *src_ptr, *dst_ptr;
-
do {
- src_ptr = kmap(*src_page) + src;
- dst_ptr = kmap(*dst_page) + dst;
+ dst_ptr = kmap_local_page(*dst_page) + dst;
+ src_ptr = kmap_local_page(*src_page) + src;
if (PAGE_SIZE - src <
PAGE_SIZE - dst) {
l = PAGE_SIZE - src;
@@ -283,9 +276,9 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
}
l = min(len, l);
memmove(dst_ptr, src_ptr, l);
- kunmap(*src_page);
+ kunmap_local(src_ptr);
set_page_dirty(*dst_page);
- kunmap(*dst_page);
+ kunmap_local(dst_ptr);
if (!dst)
dst_page++;
else
@@ -502,14 +495,14 @@ struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
if (!test_bit(HFS_BNODE_NEW, &node->flags))
return node;
- desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) +
- node->page_offset);
+ desc = (struct hfs_bnode_desc *)(kmap_local_page(node->page[0]) +
+ node->page_offset);
node->prev = be32_to_cpu(desc->prev);
node->next = be32_to_cpu(desc->next);
node->num_recs = be16_to_cpu(desc->num_recs);
node->type = desc->type;
node->height = desc->height;
- kunmap(node->page[0]);
+ kunmap_local(desc);
switch (node->type) {
case HFS_NODE_HEADER:
@@ -593,14 +586,12 @@ struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
}
pagep = node->page;
- memset(kmap(*pagep) + node->page_offset, 0,
- min_t(int, PAGE_SIZE, tree->node_size));
+ memzero_page(*pagep, node->page_offset,
+ min_t(int, PAGE_SIZE, tree->node_size));
set_page_dirty(*pagep);
- kunmap(*pagep);
for (i = 1; i < tree->pages_per_bnode; i++) {
- memset(kmap(*++pagep), 0, PAGE_SIZE);
+ memzero_page(*++pagep, 0, PAGE_SIZE);
set_page_dirty(*pagep);
- kunmap(*pagep);
}
clear_bit(HFS_BNODE_NEW, &node->flags);
wake_up(&node->lock_wq);
--
2.37.1
Hi Fabio,
> On Jul 20, 2022, at 8:43 AM, Fabio M. De Francesco <[email protected]> wrote:
>
> kmap() is being deprecated in favor of kmap_local_page().
>
> Two main problems with kmap(): (1) It comes with an overhead as mapping
> space is restricted and protected by a global lock for synchronization and
> (2) it also requires global TLB invalidation when the kmap’s pool wraps
> and it might block when the mapping space is fully utilized until a slot
> becomes available.
>
> With kmap_local_page() the mappings are per thread, CPU local, can take
> page faults, and can be called from any context (including interrupts).
> It is faster than kmap() in kernels with HIGHMEM enabled. Furthermore,
> the tasks can be preempted and, when they are scheduled to run again, the
> kernel virtual addresses are restored and still valid.
>
> Since its use in bnode.c is safe everywhere, it should be preferred.
>
> Therefore, replace kmap() with kmap_local_page() in bnode.c. Where
> possible, use the suited standard helpers (memzero_page(), memcpy_page())
> instead of open coding kmap_local_page() plus memset() or memcpy().
>
Looks reasonable. Makes sense from my point of view.
Reviewed-by: Viacheslav Dubeyko <[email protected]>
Thanks,
Slava.
> Suggested-by: Ira Weiny <[email protected]>
> Signed-off-by: Fabio M. De Francesco <[email protected]>
> ---
> fs/hfsplus/bnode.c | 105 +++++++++++++++++++++------------------------
> 1 file changed, 48 insertions(+), 57 deletions(-)
>
> diff --git a/fs/hfsplus/bnode.c b/fs/hfsplus/bnode.c
> index 177fae4e6581..3a1c77d0df48 100644
> --- a/fs/hfsplus/bnode.c
> +++ b/fs/hfsplus/bnode.c
> @@ -29,14 +29,12 @@ void hfs_bnode_read(struct hfs_bnode *node, void *buf, int off, int len)
> off &= ~PAGE_MASK;
>
> l = min_t(int, len, PAGE_SIZE - off);
> - memcpy(buf, kmap(*pagep) + off, l);
> - kunmap(*pagep);
> + memcpy_from_page(buf, *pagep, off, l);
>
> while ((len -= l) != 0) {
> buf += l;
> l = min_t(int, len, PAGE_SIZE);
> - memcpy(buf, kmap(*++pagep), l);
> - kunmap(*pagep);
> + memcpy_from_page(buf, *++pagep, 0, l);
> }
> }
>
> @@ -82,16 +80,14 @@ void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
> off &= ~PAGE_MASK;
>
> l = min_t(int, len, PAGE_SIZE - off);
> - memcpy(kmap(*pagep) + off, buf, l);
> + memcpy_to_page(*pagep, off, buf, l);
> set_page_dirty(*pagep);
> - kunmap(*pagep);
>
> while ((len -= l) != 0) {
> buf += l;
> l = min_t(int, len, PAGE_SIZE);
> - memcpy(kmap(*++pagep), buf, l);
> + memcpy_to_page(*++pagep, 0, buf, l);
> set_page_dirty(*pagep);
> - kunmap(*pagep);
> }
> }
>
> @@ -112,15 +108,13 @@ void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
> off &= ~PAGE_MASK;
>
> l = min_t(int, len, PAGE_SIZE - off);
> - memset(kmap(*pagep) + off, 0, l);
> + memzero_page(*pagep, off, l);
> set_page_dirty(*pagep);
> - kunmap(*pagep);
>
> while ((len -= l) != 0) {
> l = min_t(int, len, PAGE_SIZE);
> - memset(kmap(*++pagep), 0, l);
> + memzero_page(*++pagep, 0, l);
> set_page_dirty(*pagep);
> - kunmap(*pagep);
> }
> }
>
> @@ -142,24 +136,20 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
>
> if (src == dst) {
> l = min_t(int, len, PAGE_SIZE - src);
> - memcpy(kmap(*dst_page) + src, kmap(*src_page) + src, l);
> - kunmap(*src_page);
> + memcpy_page(*dst_page, src, *src_page, src, l);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
>
> while ((len -= l) != 0) {
> l = min_t(int, len, PAGE_SIZE);
> - memcpy(kmap(*++dst_page), kmap(*++src_page), l);
> - kunmap(*src_page);
> + memcpy_page(*++dst_page, 0, *++src_page, 0, l);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> }
> } else {
> void *src_ptr, *dst_ptr;
>
> do {
> - src_ptr = kmap(*src_page) + src;
> - dst_ptr = kmap(*dst_page) + dst;
> + dst_ptr = kmap_local_page(*dst_page) + dst;
> + src_ptr = kmap_local_page(*src_page) + src;
> if (PAGE_SIZE - src < PAGE_SIZE - dst) {
> l = PAGE_SIZE - src;
> src = 0;
> @@ -171,9 +161,9 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
> }
> l = min(len, l);
> memcpy(dst_ptr, src_ptr, l);
> - kunmap(*src_page);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
> if (!dst)
> dst_page++;
> else
> @@ -185,6 +175,7 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
> void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
> {
> struct page **src_page, **dst_page;
> + void *src_ptr, *dst_ptr;
> int l;
>
> hfs_dbg(BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len);
> @@ -202,27 +193,28 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
>
> if (src == dst) {
> while (src < len) {
> - memmove(kmap(*dst_page), kmap(*src_page), src);
> - kunmap(*src_page);
> + dst_ptr = kmap_local_page(*dst_page);
> + src_ptr = kmap_local_page(*src_page);
> + memmove(dst_ptr, src_ptr, src);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
> len -= src;
> src = PAGE_SIZE;
> src_page--;
> dst_page--;
> }
> src -= len;
> - memmove(kmap(*dst_page) + src,
> - kmap(*src_page) + src, len);
> - kunmap(*src_page);
> + dst_ptr = kmap_local_page(*dst_page);
> + src_ptr = kmap_local_page(*src_page);
> + memmove(dst_ptr + src, src_ptr + src, len);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
> } else {
> - void *src_ptr, *dst_ptr;
> -
> do {
> - src_ptr = kmap(*src_page) + src;
> - dst_ptr = kmap(*dst_page) + dst;
> + dst_ptr = kmap_local_page(*dst_page) + dst;
> + src_ptr = kmap_local_page(*src_page) + src;
> if (src < dst) {
> l = src;
> src = PAGE_SIZE;
> @@ -234,9 +226,9 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
> }
> l = min(len, l);
> memmove(dst_ptr - l, src_ptr - l, l);
> - kunmap(*src_page);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
> if (dst == PAGE_SIZE)
> dst_page--;
> else
> @@ -251,26 +243,27 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
>
> if (src == dst) {
> l = min_t(int, len, PAGE_SIZE - src);
> - memmove(kmap(*dst_page) + src,
> - kmap(*src_page) + src, l);
> - kunmap(*src_page);
> +
> + dst_ptr = kmap_local_page(*dst_page) + src;
> + src_ptr = kmap_local_page(*src_page) + src;
> + memmove(dst_ptr, src_ptr, l);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
>
> while ((len -= l) != 0) {
> l = min_t(int, len, PAGE_SIZE);
> - memmove(kmap(*++dst_page),
> - kmap(*++src_page), l);
> - kunmap(*src_page);
> + dst_ptr = kmap_local_page(*++dst_page);
> + src_ptr = kmap_local_page(*++src_page);
> + memmove(dst_ptr, src_ptr, l);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
> }
> } else {
> - void *src_ptr, *dst_ptr;
> -
> do {
> - src_ptr = kmap(*src_page) + src;
> - dst_ptr = kmap(*dst_page) + dst;
> + dst_ptr = kmap_local_page(*dst_page) + dst;
> + src_ptr = kmap_local_page(*src_page) + src;
> if (PAGE_SIZE - src <
> PAGE_SIZE - dst) {
> l = PAGE_SIZE - src;
> @@ -283,9 +276,9 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
> }
> l = min(len, l);
> memmove(dst_ptr, src_ptr, l);
> - kunmap(*src_page);
> + kunmap_local(src_ptr);
> set_page_dirty(*dst_page);
> - kunmap(*dst_page);
> + kunmap_local(dst_ptr);
> if (!dst)
> dst_page++;
> else
> @@ -502,14 +495,14 @@ struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
> if (!test_bit(HFS_BNODE_NEW, &node->flags))
> return node;
>
> - desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) +
> - node->page_offset);
> + desc = (struct hfs_bnode_desc *)(kmap_local_page(node->page[0]) +
> + node->page_offset);
> node->prev = be32_to_cpu(desc->prev);
> node->next = be32_to_cpu(desc->next);
> node->num_recs = be16_to_cpu(desc->num_recs);
> node->type = desc->type;
> node->height = desc->height;
> - kunmap(node->page[0]);
> + kunmap_local(desc);
>
> switch (node->type) {
> case HFS_NODE_HEADER:
> @@ -593,14 +586,12 @@ struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
> }
>
> pagep = node->page;
> - memset(kmap(*pagep) + node->page_offset, 0,
> - min_t(int, PAGE_SIZE, tree->node_size));
> + memzero_page(*pagep, node->page_offset,
> + min_t(int, PAGE_SIZE, tree->node_size));
> set_page_dirty(*pagep);
> - kunmap(*pagep);
> for (i = 1; i < tree->pages_per_bnode; i++) {
> - memset(kmap(*++pagep), 0, PAGE_SIZE);
> + memzero_page(*++pagep, 0, PAGE_SIZE);
> set_page_dirty(*pagep);
> - kunmap(*pagep);
> }
> clear_bit(HFS_BNODE_NEW, &node->flags);
> wake_up(&node->lock_wq);
> --
> 2.37.1
>
On mercoledì 20 luglio 2022 17:43:24 CEST Fabio M. De Francesco wrote:
> kmap() is being deprecated in favor of kmap_local_page().
>
> Two main problems with kmap(): (1) It comes with an overhead as mapping
> space is restricted and protected by a global lock for synchronization
and
> (2) it also requires global TLB invalidation when the kmap’s pool wraps
> and it might block when the mapping space is fully utilized until a slot
> becomes available.
>
> With kmap_local_page() the mappings are per thread, CPU local, can take
> page faults, and can be called from any context (including interrupts).
> It is faster than kmap() in kernels with HIGHMEM enabled. Furthermore,
> the tasks can be preempted and, when they are scheduled to run again, the
> kernel virtual addresses are restored and still valid.
>
> Since its use in bnode.c is safe everywhere, it should be preferred.
>
> Therefore, replace kmap() with kmap_local_page() in bnode.c. Where
> possible, use the suited standard helpers (memzero_page(), memcpy_page())
> instead of open coding kmap_local_page() plus memset() or memcpy().
>
> Suggested-by: Ira Weiny <[email protected]>
> Signed-off-by: Fabio M. De Francesco <[email protected]>
> ---
> fs/hfsplus/bnode.c | 105 +++++++++++++++++++++------------------------
> 1 file changed, 48 insertions(+), 57 deletions(-)
Please drop this patch because I'll send it again in a series of three, one
per file (bnode.c, bitmap.c, btree.c).
Unfortunately, while working at the last patch (btree.c), I just noticed
that I had missed one of the call sites of kmap() in bitmap.c.
Therefore, I'd prefer to send that again in a patchset and find a way to
test each of the three of them on a QEMU/KVM x86_32 VM, booting a kernel
with HIGHMEM64GB enabled.
Thanks,
Fabio