io_uring/net: Avoid msghdr on op_connect/op_bind async data
Subsystems (2)
PATCH 1/1 io_uring/net: Avoid msghdr on op_connect/op_bind async data
[ Upstream commit 3979840cd858f30f43ea9f4e7f7f1f56de82d698 ] This fixes a memory leak due to the lack of the cleanup hook for the iovec. The stable backport differs from upstream by dropping the io_connect_bpf_populate hunk, which didn't exist at the time and by fixing the merge conflict due to the introduction of io_bind_file_create.
Both IORING_OP_CONNECT and IORING_OP_BIND reuse the msghdr object just to store the sockaddr. Beyond allocating a much larger object than needed, msghdr can also wrap an iovec, which will be recycled unnecessarily. This uses the sockaddr directly.
--- io_uring/net.c | 36 ++++++++++++++++++------------------ io_uring/opdef.c | 4 ++-- 2 files changed, 20 insertions(+), 20 deletions(-)
io_uring/net.c +18 -18
@@ -1770,7 +1770,7 @@ int io_socket(struct io_kiocb *req, unsigned int issue_flags)int io_connect_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe) { struct io_connect *conn = io_kiocb_to_cmd(req, struct io_connect); -struct io_async_msghdr *io; +struct sockaddr_storage *addr;if (sqe->len || sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in)return -EINVAL ;
@@ -1779,17 +1779,17 @@ int io_connect_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) conn->addr_len =READ_ONCE (sqe->addr2); conn->in_progress = conn->seen_econnaborted = false; - io =io_msg_alloc_async (req); -if (unlikely (!io)) + addr =io_uring_alloc_async_data (NULL , req); +if (unlikely (!addr))return -ENOMEM ; -return move_addr_to_kernel (conn->addr, conn->addr_len, &io->addr); +return move_addr_to_kernel (conn->addr, conn->addr_len, addr); }int io_connect (struct io_kiocb *req,unsigned int issue_flags) { struct io_connect *connect = io_kiocb_to_cmd(req, struct io_connect); -struct io_async_msghdr *io = req->async_data; +struct sockaddr_storage *addr = req->async_data;unsigned file_flags;int ret;bool force_nonblock = issue_flags &IO_URING_F_NONBLOCK ;
@@ -1803,8 +1803,7 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags) file_flags = force_nonblock ?O_NONBLOCK :0 ; - ret =__sys_connect_file (req->file, &io->addr, connect->addr_len, - file_flags); + ret =__sys_connect_file (req->file, addr, connect->addr_len, file_flags);if ((ret == -EAGAIN || ret == -EINPROGRESS || ret == -ECONNABORTED ) && force_nonblock) {if (ret == -EINPROGRESS ) {
@@ -1833,7 +1832,6 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags) out:if (ret <0 )req_set_fail (req); -io_req_msg_cleanup (req, issue_flags);io_req_set_res (req, ret,0 );return IOU_COMPLETE ; }
@@ -1843,15 +1841,15 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags)* which in turn end up in mnt_want_write() which will grab the fs * percpu start write sem. This can trigger a lockdep warning. */ -static int io_bind_file_create (const struct io_async_msghdr *io,int addr_len) +static int io_bind_file_create (const struct sockaddr_storage *addr,int addr_len) {const struct sockaddr_un *sun; -if (io->addr.ss_family !=AF_UNIX ) +if (addr->ss_family !=AF_UNIX )return 0 ;if (addr_len <= offsetof(struct sockaddr_un , sun_path))return 0 ; - sun = (const struct sockaddr_un *) &io->addr; + sun = (const struct sockaddr_un *) addr;return sun->sun_path[0 ] !='\0' ; }
@@ -1859,7 +1857,7 @@ int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) { struct io_bind *bind = io_kiocb_to_cmd(req, struct io_bind);struct sockaddr __user *uaddr; -struct io_async_msghdr *io; +struct sockaddr_storage *addr;int ret;if (sqe->len || sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in)
@@ -1868,21 +1866,23 @@ int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) uaddr =u64_to_user_ptr (READ_ONCE (sqe->addr)); bind->addr_len =READ_ONCE (sqe->addr2); - io =io_msg_alloc_async (req); -if (unlikely (!io)) + addr =io_uring_alloc_async_data (NULL , req); +if (unlikely (!addr))return -ENOMEM ; - ret =move_addr_to_kernel (uaddr, bind->addr_len, &io->addr); + + ret =move_addr_to_kernel (uaddr, bind->addr_len, addr);if (unlikely (ret))return ret; -if (io_bind_file_create (io, bind->addr_len)) +if (io_bind_file_create (addr, bind->addr_len)) req->flags |=REQ_F_FORCE_ASYNC ;return 0 ; } +int io_bind (struct io_kiocb *req,unsigned int issue_flags) { struct io_bind *bind = io_kiocb_to_cmd(req, struct io_bind); -struct io_async_msghdr *io = req->async_data; +struct sockaddr_storage *addr = req->async_data;struct socket *sock;int ret;
@@ -1890,7 +1890,7 @@ int io_bind(struct io_kiocb *req, unsigned int issue_flags)if (unlikely (!sock))return -ENOTSOCK ; - ret =__sys_bind_socket (sock, &io->addr, bind->addr_len); + ret =__sys_bind_socket (sock, addr, bind->addr_len);if (ret <0 )req_set_fail (req);io_req_set_res (req, ret,0 );
io_uring/opdef.c +2 -2
@@ -207,7 +207,7 @@ const struct io_issue_def io_issue_defs[] = { .unbound_nonreg_file = 1, .pollout = 1,#if defined(CONFIG_NET ) - .async_size = sizeof(struct io_async_msghdr), + .async_size = sizeof(struct sockaddr_storage), .prep = io_connect_prep, .issue = io_connect,#else
@@ -504,7 +504,7 @@ const struct io_issue_def io_issue_defs[] = { .needs_file = 1, .prep = io_bind_prep, .issue = io_bind, - .async_size = sizeof(struct io_async_msghdr), + .async_size = sizeof(struct sockaddr_storage), #else .prep = io_eopnotsupp_prep,#endif
Signature · 1 line
2.54.0
Greg KH
maintainerQuoted 17 lines · Gabriel Krisman Bertazi
[ Upstream commit
3979840cd858f30f43ea9f4e7f7f1f56de82d698] This fixes a memory leak due to the lack of the cleanup hook for the iovec. The stable backport differs from upstream by dropping theio_connect_bpf_populatehunk, which didn't exist at the time and by fixing the merge conflict due to the introduction ofio_bind_file_create.Both
IORING_OP_CONNECTandIORING_OP_BINDreuse the msghdr object just to store the sockaddr. Beyond allocating a much larger object than needed, msghdr can also wrap an iovec, which will be recycled unnecessarily. This uses the sockaddr directly.Signed-off-by Gabriel Krisman Bertazi <[email protected]>Signed-off-by Jens Axboe <[email protected]>Signed-off-by Gabriel Krisman Bertazi <[email protected]>
Quoted 4 lines
--- io_uring/net.c | 36 ++++++++++++++++++------------------ io_uring/opdef.c | 4 ++-- 2 files changed, 20 insertions(+), 20 deletions(-)
This isn't in any release yet, why just 6.18? And why wan't it originally tagged for stable?
thanks,
greg k-h
Quoted 23 lines · Greg KH
Quoted 17 lines · Gabriel Krisman Bertazi
[ Upstream commit
3979840cd858f30f43ea9f4e7f7f1f56de82d698] This fixes a memory leak due to the lack of the cleanup hook for the iovec. The stable backport differs from upstream by dropping theio_connect_bpf_populatehunk, which didn't exist at the time and by fixing the merge conflict due to the introduction ofio_bind_file_create.Both
IORING_OP_CONNECTandIORING_OP_BINDreuse the msghdr object just to store the sockaddr. Beyond allocating a much larger object than needed, msghdr can also wrap an iovec, which will be recycled unnecessarily. This uses the sockaddr directly.Signed-off-by Gabriel Krisman Bertazi <[email protected]>Signed-off-by Jens Axboe <[email protected]>Signed-off-by Gabriel Krisman Bertazi <[email protected]>Quoted 4 lines
--- io_uring/net.c | 36 ++++++++++++++++++------------------ io_uring/opdef.c | 4 ++-- 2 files changed, 20 insertions(+), 20 deletions(-)
Quoted 1 line
This isn't in any release yet?
It is queued in Linus tree during the current merge window for 7.2
Quoted 1 line
why just 6.18?
The backports are slightly different, so they were sent separately. The bug exists since 6.12.
Quoted 2 lines
And why wan't it
originally tagged for stable?
Because it was originally a clean up that we later realized fixes a bug and should go to stable.
Quoted 3 lines
thanks,
greg k-h
Signature · 1 line
Gabriel Krisman Bertazi
Quoted 2 lines · Gabriel Krisman Bertazi
The backports are slightly different, so they were sent separately. The bug exists since 6.12.
6.12: https://lore.kernel.org/stable/[email protected]/T/#u
7.1: https://lore.kernel.org/stable/[email protected]/T/#u
Do you need 7.0 too? I assumed 7.0 was EOL after the 7.1 release, if not LTS.
Signature · 1 line
Gabriel Krisman Bertazi
Greg KH
maintainerQuoted 9 lines · Gabriel Krisman Bertazi
Quoted 2 lines · Gabriel Krisman Bertazi
The backports are slightly different, so they were sent separately. The bug exists since 6.12.
6.12: https://lore.kernel.org/stable/[email protected]/T/#u
7.1: https://lore.kernel.org/stable/[email protected]/T/#uDo you need 7.0 too? I assumed 7.0 was EOL after the 7.1 release, if not LTS.
It will be EOL in a week or so, it's your call. But I can't take a 6.18 or 6.12 version until that happens.
thanks,
greg k-h
PATCH 1/1 io_uring/net: Avoid msghdr on op_connect/op_bind async data
[ Upstream commit 3979840cd858f30f43ea9f4e7f7f1f56de82d698 ] This fixes a memory leak due to the lack of the cleanup hook for the iovec. The stable backport differs from upstream by dropping the io_connect_bpf_populate hunk, which didn't exist at the time and by fixing the merge conflict due to the introduction of io_bind_file_create.
Both IORING_OP_CONNECT and IORING_OP_BIND reuse the msghdr object just to store the sockaddr. Beyond allocating a much larger object than needed, msghdr can also wrap an iovec, which will be recycled unnecessarily. This uses the sockaddr directly.
--- io_uring/net.c | 36 ++++++++++++++++++------------------ io_uring/opdef.c | 4 ++-- 2 files changed, 20 insertions(+), 20 deletions(-)
io_uring/net.c +18 -18
@@ -1774,7 +1774,7 @@ int io_socket(struct io_kiocb *req, unsigned int issue_flags)int io_connect_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe) { struct io_connect *conn = io_kiocb_to_cmd(req, struct io_connect); -struct io_async_msghdr *io; +struct sockaddr_storage *addr;if (sqe->len || sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in)return -EINVAL ;
@@ -1783,17 +1783,17 @@ int io_connect_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) conn->addr_len =READ_ONCE (sqe->addr2); conn->in_progress = conn->seen_econnaborted = false; - io =io_msg_alloc_async (req); -if (unlikely (!io)) + addr =io_uring_alloc_async_data (NULL , req); +if (unlikely (!addr))return -ENOMEM ; -return move_addr_to_kernel (conn->addr, conn->addr_len, &io->addr); +return move_addr_to_kernel (conn->addr, conn->addr_len, addr); }int io_connect (struct io_kiocb *req,unsigned int issue_flags) { struct io_connect *connect = io_kiocb_to_cmd(req, struct io_connect); -struct io_async_msghdr *io = req->async_data; +struct sockaddr_storage *addr = req->async_data;unsigned file_flags;int ret;bool force_nonblock = issue_flags &IO_URING_F_NONBLOCK ;
@@ -1807,8 +1807,7 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags) file_flags = force_nonblock ?O_NONBLOCK :0 ; - ret =__sys_connect_file (req->file, &io->addr, connect->addr_len, - file_flags); + ret =__sys_connect_file (req->file, addr, connect->addr_len, file_flags);if ((ret == -EAGAIN || ret == -EINPROGRESS || ret == -ECONNABORTED ) && force_nonblock) {if (ret == -EINPROGRESS ) {
@@ -1837,7 +1836,6 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags) out:if (ret <0 )req_set_fail (req); -io_req_msg_cleanup (req, issue_flags);io_req_set_res (req, ret,0 );return IOU_COMPLETE ; }
@@ -1847,15 +1845,15 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags)* which in turn end up in mnt_want_write() which will grab the fs * percpu start write sem. This can trigger a lockdep warning. */ -static int io_bind_file_create (const struct io_async_msghdr *io,int addr_len) +static int io_bind_file_create (const struct sockaddr_storage *addr,int addr_len) {const struct sockaddr_un *sun; -if (io->addr.ss_family !=AF_UNIX ) +if (addr->ss_family !=AF_UNIX )return 0 ;if (addr_len <= offsetof(struct sockaddr_un , sun_path))return 0 ; - sun = (const struct sockaddr_un *) &io->addr; + sun = (const struct sockaddr_un *) addr;return sun->sun_path[0 ] !='\0' ; }
@@ -1863,7 +1861,7 @@ int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) { struct io_bind *bind = io_kiocb_to_cmd(req, struct io_bind);struct sockaddr __user *uaddr; -struct io_async_msghdr *io; +struct sockaddr_storage *addr;int ret;if (sqe->len || sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in)
@@ -1872,21 +1870,23 @@ int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) uaddr =u64_to_user_ptr (READ_ONCE (sqe->addr)); bind->addr_len =READ_ONCE (sqe->addr2); - io =io_msg_alloc_async (req); -if (unlikely (!io)) + addr =io_uring_alloc_async_data (NULL , req); +if (unlikely (!addr))return -ENOMEM ; - ret =move_addr_to_kernel (uaddr, bind->addr_len, &io->addr); + + ret =move_addr_to_kernel (uaddr, bind->addr_len, addr);if (unlikely (ret))return ret; -if (io_bind_file_create (io, bind->addr_len)) +if (io_bind_file_create (addr, bind->addr_len)) req->flags |=REQ_F_FORCE_ASYNC ;return 0 ; } +int io_bind (struct io_kiocb *req,unsigned int issue_flags) { struct io_bind *bind = io_kiocb_to_cmd(req, struct io_bind); -struct io_async_msghdr *io = req->async_data; +struct sockaddr_storage *addr = req->async_data;struct socket *sock;int ret;
@@ -1894,7 +1894,7 @@ int io_bind(struct io_kiocb *req, unsigned int issue_flags)if (unlikely (!sock))return -ENOTSOCK ; - ret =__sys_bind_socket (sock, &io->addr, bind->addr_len); + ret =__sys_bind_socket (sock, addr, bind->addr_len);if (ret <0 )req_set_fail (req);io_req_set_res (req, ret,0 );
io_uring/opdef.c +2 -2
@@ -207,7 +207,7 @@ const struct io_issue_def io_issue_defs[] = { .unbound_nonreg_file = 1, .pollout = 1,#if defined(CONFIG_NET ) - .async_size = sizeof(struct io_async_msghdr), + .async_size = sizeof(struct sockaddr_storage), .prep = io_connect_prep, .issue = io_connect,#else
@@ -510,7 +510,7 @@ const struct io_issue_def io_issue_defs[] = { .needs_file = 1, .prep = io_bind_prep, .issue = io_bind, - .async_size = sizeof(struct io_async_msghdr), + .async_size = sizeof(struct sockaddr_storage), #else .prep = io_eopnotsupp_prep,#endif
Signature · 1 line
2.54.0
Quoted 2 lines
[PATCH stable-7.0.y]
io_uring/net: Avoid msghdr on op_connect/op_bind async data
Queued for 7.0, thanks.
--
Thanks,
Sasha
Greg KH
maintainerQuoted 25 lines · Gabriel Krisman Bertazi
Quoted 23 lines · Greg KH
Quoted 21 lines · Gabriel Krisman Bertazi
[ Upstream commit
3979840cd858f30f43ea9f4e7f7f1f56de82d698] This fixes a memory leak due to the lack of the cleanup hook for the iovec. The stable backport differs from upstream by dropping theio_connect_bpf_populatehunk, which didn't exist at the time and by fixing the merge conflict due to the introduction ofio_bind_file_create.Both
IORING_OP_CONNECTandIORING_OP_BINDreuse the msghdr object just to store the sockaddr. Beyond allocating a much larger object than needed, msghdr can also wrap an iovec, which will be recycled unnecessarily. This uses the sockaddr directly.Signed-off-by Gabriel Krisman Bertazi <[email protected]>Signed-off-by Jens Axboe <[email protected]>Signed-off-by Gabriel Krisman Bertazi <[email protected]>--- io_uring/net.c | 36 ++++++++++++++++++------------------ io_uring/opdef.c | 4 ++-- 2 files changed, 20 insertions(+), 20 deletions(-)
Quoted 3 lines
Quoted 1 line
This isn't in any release yet?
It is queued in Linus tree during the current merge window for 7.2
Ah, so it's not even in a released -rc yet.
Quoted 4 lines
Quoted 1 line
why just 6.18?
The backports are slightly different, so they were sent separately. The bug exists since 6.12.
I missed seeing a backport for 7.0, shouldn't it also go there?
thanks,
greg k-h