summaryrefslogtreecommitdiffstats
path: root/io_uring/kbuf.c
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-05-18 18:50:03 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-05-18 18:50:03 +0000
commit01a69402cf9d38ff180345d55c2ee51c7e89fbc7 (patch)
treeb406c5242a088c4f59c6e4b719b783f43aca6ae9 /io_uring/kbuf.c
parentAdding upstream version 6.7.12. (diff)
downloadlinux-01a69402cf9d38ff180345d55c2ee51c7e89fbc7.tar.xz
linux-01a69402cf9d38ff180345d55c2ee51c7e89fbc7.zip
Adding upstream version 6.8.9.upstream/6.8.9
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'io_uring/kbuf.c')
-rw-r--r--io_uring/kbuf.c142
1 files changed, 66 insertions, 76 deletions
diff --git a/io_uring/kbuf.c b/io_uring/kbuf.c
index 72b6af1d2e..b2c4f82937 100644
--- a/io_uring/kbuf.c
+++ b/io_uring/kbuf.c
@@ -17,8 +17,6 @@
#define IO_BUFFER_LIST_BUF_PER_PAGE (PAGE_SIZE / sizeof(struct io_uring_buf))
-#define BGID_ARRAY 64
-
/* BIDs are addressed by a 16-bit field in a CQE */
#define MAX_BIDS_PER_BGID (1 << 16)
@@ -40,13 +38,9 @@ struct io_buf_free {
int inuse;
};
-static struct io_buffer_list *__io_buffer_get_list(struct io_ring_ctx *ctx,
- struct io_buffer_list *bl,
- unsigned int bgid)
+static inline struct io_buffer_list *__io_buffer_get_list(struct io_ring_ctx *ctx,
+ unsigned int bgid)
{
- if (bl && bgid < BGID_ARRAY)
- return &bl[bgid];
-
return xa_load(&ctx->io_bl_xa, bgid);
}
@@ -55,7 +49,7 @@ static inline struct io_buffer_list *io_buffer_get_list(struct io_ring_ctx *ctx,
{
lockdep_assert_held(&ctx->uring_lock);
- return __io_buffer_get_list(ctx, ctx->io_bl, bgid);
+ return __io_buffer_get_list(ctx, bgid);
}
static int io_buffer_add_list(struct io_ring_ctx *ctx,
@@ -67,11 +61,7 @@ static int io_buffer_add_list(struct io_ring_ctx *ctx,
* always under the ->uring_lock, but the RCU lookup from mmap does.
*/
bl->bgid = bgid;
- smp_store_release(&bl->is_ready, 1);
-
- if (bgid < BGID_ARRAY)
- return 0;
-
+ atomic_set(&bl->refs, 1);
return xa_err(xa_store(&ctx->io_bl_xa, bgid, bl, GFP_KERNEL));
}
@@ -217,24 +207,6 @@ void __user *io_buffer_select(struct io_kiocb *req, size_t *len,
return ret;
}
-static __cold int io_init_bl_list(struct io_ring_ctx *ctx)
-{
- struct io_buffer_list *bl;
- int i;
-
- bl = kcalloc(BGID_ARRAY, sizeof(struct io_buffer_list), GFP_KERNEL);
- if (!bl)
- return -ENOMEM;
-
- for (i = 0; i < BGID_ARRAY; i++) {
- INIT_LIST_HEAD(&bl[i].buf_list);
- bl[i].bgid = i;
- }
-
- smp_store_release(&ctx->io_bl, bl);
- return 0;
-}
-
/*
* Mark the given mapped range as free for reuse
*/
@@ -303,24 +275,24 @@ static int __io_remove_buffers(struct io_ring_ctx *ctx,
return i;
}
+void io_put_bl(struct io_ring_ctx *ctx, struct io_buffer_list *bl)
+{
+ if (atomic_dec_and_test(&bl->refs)) {
+ __io_remove_buffers(ctx, bl, -1U);
+ kfree_rcu(bl, rcu);
+ }
+}
+
void io_destroy_buffers(struct io_ring_ctx *ctx)
{
struct io_buffer_list *bl;
struct list_head *item, *tmp;
struct io_buffer *buf;
unsigned long index;
- int i;
-
- for (i = 0; i < BGID_ARRAY; i++) {
- if (!ctx->io_bl)
- break;
- __io_remove_buffers(ctx, &ctx->io_bl[i], -1U);
- }
xa_for_each(&ctx->io_bl_xa, index, bl) {
xa_erase(&ctx->io_bl_xa, bl->bgid);
- __io_remove_buffers(ctx, bl, -1U);
- kfree_rcu(bl, rcu);
+ io_put_bl(ctx, bl);
}
/*
@@ -498,12 +470,6 @@ int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags)
io_ring_submit_lock(ctx, issue_flags);
- if (unlikely(p->bgid < BGID_ARRAY && !ctx->io_bl)) {
- ret = io_init_bl_list(ctx);
- if (ret)
- goto err;
- }
-
bl = io_buffer_get_list(ctx, p->bgid);
if (unlikely(!bl)) {
bl = kzalloc(sizeof(*bl), GFP_KERNEL_ACCOUNT);
@@ -516,14 +482,9 @@ int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags)
if (ret) {
/*
* Doesn't need rcu free as it was never visible, but
- * let's keep it consistent throughout. Also can't
- * be a lower indexed array group, as adding one
- * where lookup failed cannot happen.
+ * let's keep it consistent throughout.
*/
- if (p->bgid >= BGID_ARRAY)
- kfree_rcu(bl, rcu);
- else
- WARN_ON_ONCE(1);
+ kfree_rcu(bl, rcu);
goto err;
}
}
@@ -688,12 +649,6 @@ int io_register_pbuf_ring(struct io_ring_ctx *ctx, void __user *arg)
if (reg.ring_entries >= 65536)
return -EINVAL;
- if (unlikely(reg.bgid < BGID_ARRAY && !ctx->io_bl)) {
- int ret = io_init_bl_list(ctx);
- if (ret)
- return ret;
- }
-
bl = io_buffer_get_list(ctx, reg.bgid);
if (bl) {
/* if mapped buffer ring OR classic exists, don't allow */
@@ -742,31 +697,66 @@ int io_unregister_pbuf_ring(struct io_ring_ctx *ctx, void __user *arg)
if (!bl->is_mapped)
return -EINVAL;
- __io_remove_buffers(ctx, bl, -1U);
- if (bl->bgid >= BGID_ARRAY) {
- xa_erase(&ctx->io_bl_xa, bl->bgid);
- kfree_rcu(bl, rcu);
- }
+ xa_erase(&ctx->io_bl_xa, bl->bgid);
+ io_put_bl(ctx, bl);
return 0;
}
-void *io_pbuf_get_address(struct io_ring_ctx *ctx, unsigned long bgid)
+int io_register_pbuf_status(struct io_ring_ctx *ctx, void __user *arg)
{
+ struct io_uring_buf_status buf_status;
struct io_buffer_list *bl;
+ int i;
- bl = __io_buffer_get_list(ctx, smp_load_acquire(&ctx->io_bl), bgid);
+ if (copy_from_user(&buf_status, arg, sizeof(buf_status)))
+ return -EFAULT;
+
+ for (i = 0; i < ARRAY_SIZE(buf_status.resv); i++)
+ if (buf_status.resv[i])
+ return -EINVAL;
+
+ bl = io_buffer_get_list(ctx, buf_status.buf_group);
+ if (!bl)
+ return -ENOENT;
+ if (!bl->is_mapped)
+ return -EINVAL;
+
+ buf_status.head = bl->head;
+ if (copy_to_user(arg, &buf_status, sizeof(buf_status)))
+ return -EFAULT;
+
+ return 0;
+}
+
+struct io_buffer_list *io_pbuf_get_bl(struct io_ring_ctx *ctx,
+ unsigned long bgid)
+{
+ struct io_buffer_list *bl;
+ bool ret;
- if (!bl || !bl->is_mmap)
- return NULL;
/*
- * Ensure the list is fully setup. Only strictly needed for RCU lookup
- * via mmap, and in that case only for the array indexed groups. For
- * the xarray lookups, it's either visible and ready, or not at all.
+ * We have to be a bit careful here - we're inside mmap and cannot grab
+ * the uring_lock. This means the buffer_list could be simultaneously
+ * going away, if someone is trying to be sneaky. Look it up under rcu
+ * so we know it's not going away, and attempt to grab a reference to
+ * it. If the ref is already zero, then fail the mapping. If successful,
+ * the caller will call io_put_bl() to drop the the reference at at the
+ * end. This may then safely free the buffer_list (and drop the pages)
+ * at that point, vm_insert_pages() would've already grabbed the
+ * necessary vma references.
*/
- if (!smp_load_acquire(&bl->is_ready))
- return NULL;
-
- return bl->buf_ring;
+ rcu_read_lock();
+ bl = xa_load(&ctx->io_bl_xa, bgid);
+ /* must be a mmap'able buffer ring and have pages */
+ ret = false;
+ if (bl && bl->is_mmap)
+ ret = atomic_inc_not_zero(&bl->refs);
+ rcu_read_unlock();
+
+ if (ret)
+ return bl;
+
+ return ERR_PTR(-EINVAL);
}
/*