[PATCH 3/7] io_uring/bpf_filter: allow filtering on contents of struct open_how

[PATCH 3/7] io_uring/bpf_filter: allow filtering on contents of struct open_how

[Jens Axboe]

This adds custom filtering for IORING_OP_OPENAT and IORING_OP_OPENAT2,
where the open_how flags, mode, and resolve can be checked by filters.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/uapi/linux/io_uring/bpf_filter.h | 5 +++++
 io_uring/bpf_filter.c                    | 5 +++++
 io_uring/openclose.c                     | 9 +++++++++
 io_uring/openclose.h                     | 3 +++
 4 files changed, 22 insertions(+)

diff --git a/include/uapi/linux/io_uring/bpf_filter.h b/include/uapi/linux/io_uring/bpf_filter.h
index ad6961be5efa..7f468628c491 100644
--- a/include/uapi/linux/io_uring/bpf_filter.h
+++ b/include/uapi/linux/io_uring/bpf_filter.h
@@ -22,6 +22,11 @@ struct io_uring_bpf_ctx {
 			__u32	type;
 			__u32	protocol;
 		} socket;
+		struct {
+			__u64	flags;
+			__u64	mode;
+			__u64	resolve;
+		} open;
 	};
 };
 
diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
index 8934c0586842..3352f53fd2b9 100644
--- a/io_uring/bpf_filter.c
+++ b/io_uring/bpf_filter.c
@@ -12,6 +12,7 @@
 #include "io_uring.h"
 #include "bpf_filter.h"
 #include "net.h"
+#include "openclose.h"
 
 struct io_bpf_filter {
 	struct bpf_prog		*prog;
@@ -38,6 +39,10 @@ static void io_uring_populate_bpf_ctx(struct io_uring_bpf_ctx *bctx,
 	case IORING_OP_SOCKET:
 		io_socket_bpf_populate(bctx, req);
 		break;
+	case IORING_OP_OPENAT:
+	case IORING_OP_OPENAT2:
+		io_openat_bpf_populate(bctx, req);
+		break;
 	}
 }
 
diff --git a/io_uring/openclose.c b/io_uring/openclose.c
index 15dde9bd6ff6..31c687adf873 100644
--- a/io_uring/openclose.c
+++ b/io_uring/openclose.c
@@ -85,6 +85,15 @@ static int __io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe
 	return 0;
 }
 
+void io_openat_bpf_populate(struct io_uring_bpf_ctx *bctx, struct io_kiocb *req)
+{
+	struct io_open *open = io_kiocb_to_cmd(req, struct io_open);
+
+	bctx->open.flags = open->how.flags;
+	bctx->open.mode = open->how.mode;
+	bctx->open.resolve = open->how.resolve;
+}
+
 int io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	struct io_open *open = io_kiocb_to_cmd(req, struct io_open);
diff --git a/io_uring/openclose.h b/io_uring/openclose.h
index 4ca2a9935abc..566739920658 100644
--- a/io_uring/openclose.h
+++ b/io_uring/openclose.h
@@ -1,11 +1,14 @@
 // SPDX-License-Identifier: GPL-2.0
 
+#include "bpf_filter.h"
+
 int __io_close_fixed(struct io_ring_ctx *ctx, unsigned int issue_flags,
 		     unsigned int offset);
 
 int io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_openat(struct io_kiocb *req, unsigned int issue_flags);
 void io_open_cleanup(struct io_kiocb *req);
+void io_openat_bpf_populate(struct io_uring_bpf_ctx *bctx, struct io_kiocb *req);
 
 int io_openat2_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_openat2(struct io_kiocb *req, unsigned int issue_flags);
-- 
2.51.0

Re: [PATCH 3/7] io_uring/bpf_filter: allow filtering on contents of struct open_how

[Christian Brauner]

On Mon, Jan 19, 2026 at 04:54:26PM -0700, Jens Axboe wrote:
> This adds custom filtering for IORING_OP_OPENAT and IORING_OP_OPENAT2,
> where the open_how flags, mode, and resolve can be checked by filters.
> 
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> ---
>  include/uapi/linux/io_uring/bpf_filter.h | 5 +++++
>  io_uring/bpf_filter.c                    | 5 +++++
>  io_uring/openclose.c                     | 9 +++++++++
>  io_uring/openclose.h                     | 3 +++
>  4 files changed, 22 insertions(+)
> 
> diff --git a/include/uapi/linux/io_uring/bpf_filter.h b/include/uapi/linux/io_uring/bpf_filter.h
> index ad6961be5efa..7f468628c491 100644
> --- a/include/uapi/linux/io_uring/bpf_filter.h
> +++ b/include/uapi/linux/io_uring/bpf_filter.h
> @@ -22,6 +22,11 @@ struct io_uring_bpf_ctx {
>  			__u32	type;
>  			__u32	protocol;
>  		} socket;
> +		struct {
> +			__u64	flags;
> +			__u64	mode;
> +			__u64	resolve;
> +		} open;

So openat2()'s struct is extensible and there are plans to extend it to
include e.g., upgrade masks to restrict how a file descriptor can be
reopened. And in general there's the potential that this gets extended
with additional fields. So if it's workable I would add a size argument
in here to communicate to the bpf program what io_uring currently knows
about/is able to filter on. That should be fairly simple and doesn't
require you to change a whole lot?

[PATCHSET v7] Inherited restrictions and BPF filtering for io_uring

[Jens Axboe]

Hi,

Followup to v6 here:

https://lore.kernel.org/io-uring/20260119235456.1722452-1-axboe@kernel.dk/

Mostly just addressing a bit of feedback, feature wise this is all the
same as before. See the changelog section for more details.

Kernel branch can be found here:

https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux.git/log/?h=io_uring-bpf-restrictions.4

and a liburing branch with support helpers, man page, and a fairly
substantial test case can be found here:

https://git.kernel.org/pub/scm/linux/kernel/git/axboe/liburing.git/log/?h=bpf-restrictions

Feedback welcome!

Changes since v6
- Add pdu_size to struct io_uring_bpf_ctx. This will help future proof
  filters as they can check if they agree with the kernel on what is
  available, and it eliminates the need to pad the struct out for future
  expansion (Christian)
- Various code cleanups (Christian)
- Fix for FORTIFY on how the bctx filter struct is cleared.
- Rebase on current for-7.0/io_uring tree.

 include/linux/io_uring.h                 |  14 +-
 include/linux/io_uring_types.h           |  13 +
 include/linux/sched.h                    |   1 +
 include/uapi/linux/io_uring.h            |  10 +
 include/uapi/linux/io_uring/bpf_filter.h |  62 ++++
 io_uring/Kconfig                         |   5 +
 io_uring/Makefile                        |   1 +
 io_uring/bpf_filter.c                    | 430 +++++++++++++++++++++++
 io_uring/bpf_filter.h                    |  48 +++
 io_uring/io_uring.c                      |  48 +++
 io_uring/io_uring.h                      |   1 +
 io_uring/net.c                           |   9 +
 io_uring/net.h                           |   6 +
 io_uring/openclose.c                     |   9 +
 io_uring/openclose.h                     |   3 +
 io_uring/register.c                      |  91 +++++
 io_uring/tctx.c                          |  42 ++-
 kernel/fork.c                            |   5 +
 18 files changed, 788 insertions(+), 10 deletions(-)

--
Jens Axboe

[PATCH 1/7] io_uring: add support for BPF filtering for opcode restrictions

[Jens Axboe]

Add support for loading classic BPF programs with io_uring to provide
fine-grained filtering of SQE operations. Unlike
IORING_REGISTER_RESTRICTIONS which only allows bitmap-based allow/deny
of opcodes, BPF filters can inspect request attributes and make dynamic
decisions.

The filter is registered via IORING_REGISTER_BPF_FILTER with a struct
io_uring_bpf:

struct io_uring_bpf_filter {
	__u32	opcode;		/* io_uring opcode to filter */
	__u32	flags;
	__u32	filter_len;	/* number of BPF instructions */
	__u32	resv;
	__u64	filter_ptr;	/* pointer to BPF filter */
	__u64	resv2[5];
};

enum {
	IO_URING_BPF_CMD_FILTER	= 1,
};

struct io_uring_bpf {
	__u16	cmd_type;	/* IO_URING_BPF_* values */
	__u16	cmd_flags;	/* none so far */
	__u32	resv;
	union {
		struct io_uring_bpf_filter	filter;
	};
};

and the filters get supplied a struct io_uring_bpf_ctx:

struct io_uring_bpf_ctx {
	__u64	user_data;
	__u8	opcode;
	__u8	sqe_flags;
	__u8	pdu_size;
	__u8	pad[5];
};

where it's possible to filter on opcode and sqe_flags, with pdu_size
indicating how much extra data is being passed in beyond the pad field.
This will used for specific finer grained filtering inside an opcode.
An example of that for sockets is in one of the following patches.
Anything the opcode supports can end up in this struct, populated by
the opcode itself, and hence can be filtered for.

Filters have the following semantics:
  - Return 1 to allow the request
  - Return 0 to deny the request with -EACCES
  - Multiple filters can be stacked per opcode. All filters must
    return 1 for the opcode to be allowed.
  - Filters are evaluated in registration order (most recent first)

The implementation uses classic BPF (cBPF) rather than eBPF for as
that's required for containers, and since they can be used by any
user in the system.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/io_uring_types.h           |   9 +
 include/uapi/linux/io_uring.h            |   3 +
 include/uapi/linux/io_uring/bpf_filter.h |  50 ++++
 io_uring/Kconfig                         |   5 +
 io_uring/Makefile                        |   1 +
 io_uring/bpf_filter.c                    | 321 +++++++++++++++++++++++
 io_uring/bpf_filter.h                    |  42 +++
 io_uring/io_uring.c                      |   8 +
 io_uring/register.c                      |   8 +
 9 files changed, 447 insertions(+)
 create mode 100644 include/uapi/linux/io_uring/bpf_filter.h
 create mode 100644 io_uring/bpf_filter.c
 create mode 100644 io_uring/bpf_filter.h

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index dc6bd6940a0d..74bf98362876 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -219,9 +219,18 @@ struct io_rings {
 	struct io_uring_cqe	cqes[] ____cacheline_aligned_in_smp;
 };
 
+struct io_bpf_filter;
+struct io_bpf_filters {
+	refcount_t refs;	/* ref for ->bpf_filters */
+	spinlock_t lock;	/* protects ->bpf_filters modifications */
+	struct io_bpf_filter __rcu **filters;
+	struct rcu_head rcu_head;
+};
+
 struct io_restriction {
 	DECLARE_BITMAP(register_op, IORING_REGISTER_LAST);
 	DECLARE_BITMAP(sqe_op, IORING_OP_LAST);
+	struct io_bpf_filters *bpf_filters;
 	u8 sqe_flags_allowed;
 	u8 sqe_flags_required;
 	/* IORING_OP_* restrictions exist */
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index b5b23c0d5283..94669b77fee8 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -700,6 +700,9 @@ enum io_uring_register_op {
 	/* auxiliary zcrx configuration, see enum zcrx_ctrl_op */
 	IORING_REGISTER_ZCRX_CTRL		= 36,
 
+	/* register bpf filtering programs */
+	IORING_REGISTER_BPF_FILTER		= 37,
+
 	/* this goes last */
 	IORING_REGISTER_LAST,
 
diff --git a/include/uapi/linux/io_uring/bpf_filter.h b/include/uapi/linux/io_uring/bpf_filter.h
new file mode 100644
index 000000000000..2d4d0e5743e4
--- /dev/null
+++ b/include/uapi/linux/io_uring/bpf_filter.h
@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: (GPL-2.0 WITH Linux-syscall-note) OR MIT */
+/*
+ * Header file for the io_uring BPF filters.
+ */
+#ifndef LINUX_IO_URING_BPF_FILTER_H
+#define LINUX_IO_URING_BPF_FILTER_H
+
+#include <linux/types.h>
+
+/*
+ * Struct passed to filters.
+ */
+struct io_uring_bpf_ctx {
+	__u64	user_data;
+	__u8	opcode;
+	__u8	sqe_flags;
+	__u8	pdu_size;	/* size of aux data for filter */
+	__u8	pad[5];
+};
+
+enum {
+	/*
+	 * If set, any currently unset opcode will have a deny filter attached
+	 */
+	IO_URING_BPF_FILTER_DENY_REST	= 1,
+};
+
+struct io_uring_bpf_filter {
+	__u32	opcode;		/* io_uring opcode to filter */
+	__u32	flags;
+	__u32	filter_len;	/* number of BPF instructions */
+	__u32	resv;
+	__u64	filter_ptr;	/* pointer to BPF filter */
+	__u64	resv2[5];
+};
+
+enum {
+	IO_URING_BPF_CMD_FILTER	= 1,
+};
+
+struct io_uring_bpf {
+	__u16	cmd_type;	/* IO_URING_BPF_* values */
+	__u16	cmd_flags;	/* none so far */
+	__u32	resv;
+	union {
+		struct io_uring_bpf_filter	filter;
+	};
+};
+
+#endif
diff --git a/io_uring/Kconfig b/io_uring/Kconfig
index 4b949c42c0bf..a7ae23cf1035 100644
--- a/io_uring/Kconfig
+++ b/io_uring/Kconfig
@@ -9,3 +9,8 @@ config IO_URING_ZCRX
 	depends on PAGE_POOL
 	depends on INET
 	depends on NET_RX_BUSY_POLL
+
+config IO_URING_BPF
+	def_bool y
+	depends on BPF
+	depends on NET
diff --git a/io_uring/Makefile b/io_uring/Makefile
index bf9eff88427a..931f9156132a 100644
--- a/io_uring/Makefile
+++ b/io_uring/Makefile
@@ -24,3 +24,4 @@ obj-$(CONFIG_NET_RX_BUSY_POLL)	+= napi.o
 obj-$(CONFIG_NET) += net.o cmd_net.o
 obj-$(CONFIG_PROC_FS) += fdinfo.o
 obj-$(CONFIG_IO_URING_MOCK_FILE) += mock_file.o
+obj-$(CONFIG_IO_URING_BPF) += bpf_filter.o
diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
new file mode 100644
index 000000000000..5207226d72ea
--- /dev/null
+++ b/io_uring/bpf_filter.c
@@ -0,0 +1,321 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * BPF filter support for io_uring. Supports SQE opcodes for now.
+ */
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/io_uring.h>
+#include <linux/filter.h>
+#include <linux/bpf.h>
+#include <uapi/linux/io_uring.h>
+
+#include "io_uring.h"
+#include "bpf_filter.h"
+#include "net.h"
+
+struct io_bpf_filter {
+	struct bpf_prog		*prog;
+	struct io_bpf_filter	*next;
+};
+
+/* Deny if this is set as the filter */
+static const struct io_bpf_filter dummy_filter;
+
+static void io_uring_populate_bpf_ctx(struct io_uring_bpf_ctx *bctx,
+				      struct io_kiocb *req)
+{
+	bctx->opcode = req->opcode;
+	bctx->sqe_flags = (__force int) req->flags & SQE_VALID_FLAGS;
+	bctx->user_data = req->cqe.user_data;
+	/* clear residual, anything from pdu_size and below */
+	memset((void *) bctx + offsetof(struct io_uring_bpf_ctx, pdu_size), 0,
+		sizeof(*bctx) - offsetof(struct io_uring_bpf_ctx, pdu_size));
+}
+
+/*
+ * Run registered filters for a given opcode. For filters, a return of 0 denies
+ * execution of the request, a return of 1 allows it. If any filter for an
+ * opcode returns 0, filter processing is stopped, and the request is denied.
+ * This also stops the processing of filters.
+ *
+ * __io_uring_run_bpf_filters() returns 0 on success, allow running the
+ * request, and -EACCES when a request is denied.
+ */
+int __io_uring_run_bpf_filters(struct io_restriction *res, struct io_kiocb *req)
+{
+	struct io_bpf_filter *filter;
+	struct io_uring_bpf_ctx bpf_ctx;
+	int ret;
+
+	/* Fast check for existence of filters outside of RCU */
+	if (!rcu_access_pointer(res->bpf_filters->filters[req->opcode]))
+		return 0;
+
+	/*
+	 * req->opcode has already been validated to be within the range
+	 * of what we expect, io_init_req() does this.
+	 */
+	guard(rcu)();
+	filter = rcu_dereference(res->bpf_filters->filters[req->opcode]);
+	if (!filter)
+		return 0;
+	else if (filter == &dummy_filter)
+		return -EACCES;
+
+	io_uring_populate_bpf_ctx(&bpf_ctx, req);
+
+	/*
+	 * Iterate registered filters. The opcode is allowed IFF all filters
+	 * return 1. If any filter returns denied, opcode will be denied.
+	 */
+	do {
+		if (filter == &dummy_filter)
+			return -EACCES;
+		ret = bpf_prog_run(filter->prog, &bpf_ctx);
+		if (!ret)
+			return -EACCES;
+		filter = filter->next;
+	} while (filter);
+
+	return 0;
+}
+
+static void io_free_bpf_filters(struct rcu_head *head)
+{
+	struct io_bpf_filter __rcu **filter;
+	struct io_bpf_filters *filters;
+	int i;
+
+	filters = container_of(head, struct io_bpf_filters, rcu_head);
+	scoped_guard(spinlock, &filters->lock) {
+		filter = filters->filters;
+		if (!filter)
+			return;
+	}
+
+	for (i = 0; i < IORING_OP_LAST; i++) {
+		struct io_bpf_filter *f;
+
+		rcu_read_lock();
+		f = rcu_dereference(filter[i]);
+		while (f) {
+			struct io_bpf_filter *next = f->next;
+
+			/*
+			 * Even if stacked, dummy filter will always be last
+			 * as it can only get installed into an empty spot.
+			 */
+			if (f == &dummy_filter)
+				break;
+			bpf_prog_destroy(f->prog);
+			kfree(f);
+			f = next;
+		}
+		rcu_read_unlock();
+	}
+	kfree(filters->filters);
+	kfree(filters);
+}
+
+static void __io_put_bpf_filters(struct io_bpf_filters *filters)
+{
+	if (refcount_dec_and_test(&filters->refs))
+		call_rcu(&filters->rcu_head, io_free_bpf_filters);
+}
+
+void io_put_bpf_filters(struct io_restriction *res)
+{
+	if (res->bpf_filters)
+		__io_put_bpf_filters(res->bpf_filters);
+}
+
+static struct io_bpf_filters *io_new_bpf_filters(void)
+{
+	struct io_bpf_filters *filters __free(kfree) = NULL;
+
+	filters = kzalloc(sizeof(*filters), GFP_KERNEL_ACCOUNT);
+	if (!filters)
+		return ERR_PTR(-ENOMEM);
+
+	filters->filters = kcalloc(IORING_OP_LAST,
+				   sizeof(struct io_bpf_filter *),
+				   GFP_KERNEL_ACCOUNT);
+	if (!filters->filters)
+		return ERR_PTR(-ENOMEM);
+
+	refcount_set(&filters->refs, 1);
+	spin_lock_init(&filters->lock);
+	return no_free_ptr(filters);
+}
+
+/*
+ * Validate classic BPF filter instructions. Only allow a safe subset of
+ * operations - no packet data access, just context field loads and basic
+ * ALU/jump operations.
+ */
+static int io_uring_check_cbpf_filter(struct sock_filter *filter,
+				      unsigned int flen)
+{
+	int pc;
+
+	for (pc = 0; pc < flen; pc++) {
+		struct sock_filter *ftest = &filter[pc];
+		u16 code = ftest->code;
+		u32 k = ftest->k;
+
+		switch (code) {
+		case BPF_LD | BPF_W | BPF_ABS:
+			ftest->code = BPF_LDX | BPF_W | BPF_ABS;
+			/* 32-bit aligned and not out of bounds. */
+			if (k >= sizeof(struct io_uring_bpf_ctx) || k & 3)
+				return -EINVAL;
+			continue;
+		case BPF_LD | BPF_W | BPF_LEN:
+			ftest->code = BPF_LD | BPF_IMM;
+			ftest->k = sizeof(struct io_uring_bpf_ctx);
+			continue;
+		case BPF_LDX | BPF_W | BPF_LEN:
+			ftest->code = BPF_LDX | BPF_IMM;
+			ftest->k = sizeof(struct io_uring_bpf_ctx);
+			continue;
+		/* Explicitly include allowed calls. */
+		case BPF_RET | BPF_K:
+		case BPF_RET | BPF_A:
+		case BPF_ALU | BPF_ADD | BPF_K:
+		case BPF_ALU | BPF_ADD | BPF_X:
+		case BPF_ALU | BPF_SUB | BPF_K:
+		case BPF_ALU | BPF_SUB | BPF_X:
+		case BPF_ALU | BPF_MUL | BPF_K:
+		case BPF_ALU | BPF_MUL | BPF_X:
+		case BPF_ALU | BPF_DIV | BPF_K:
+		case BPF_ALU | BPF_DIV | BPF_X:
+		case BPF_ALU | BPF_AND | BPF_K:
+		case BPF_ALU | BPF_AND | BPF_X:
+		case BPF_ALU | BPF_OR | BPF_K:
+		case BPF_ALU | BPF_OR | BPF_X:
+		case BPF_ALU | BPF_XOR | BPF_K:
+		case BPF_ALU | BPF_XOR | BPF_X:
+		case BPF_ALU | BPF_LSH | BPF_K:
+		case BPF_ALU | BPF_LSH | BPF_X:
+		case BPF_ALU | BPF_RSH | BPF_K:
+		case BPF_ALU | BPF_RSH | BPF_X:
+		case BPF_ALU | BPF_NEG:
+		case BPF_LD | BPF_IMM:
+		case BPF_LDX | BPF_IMM:
+		case BPF_MISC | BPF_TAX:
+		case BPF_MISC | BPF_TXA:
+		case BPF_LD | BPF_MEM:
+		case BPF_LDX | BPF_MEM:
+		case BPF_ST:
+		case BPF_STX:
+		case BPF_JMP | BPF_JA:
+		case BPF_JMP | BPF_JEQ | BPF_K:
+		case BPF_JMP | BPF_JEQ | BPF_X:
+		case BPF_JMP | BPF_JGE | BPF_K:
+		case BPF_JMP | BPF_JGE | BPF_X:
+		case BPF_JMP | BPF_JGT | BPF_K:
+		case BPF_JMP | BPF_JGT | BPF_X:
+		case BPF_JMP | BPF_JSET | BPF_K:
+		case BPF_JMP | BPF_JSET | BPF_X:
+			continue;
+		default:
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
+#define IO_URING_BPF_FILTER_FLAGS	IO_URING_BPF_FILTER_DENY_REST
+
+int io_register_bpf_filter(struct io_restriction *res,
+			   struct io_uring_bpf __user *arg)
+{
+	struct io_bpf_filter *filter, *old_filter;
+	struct io_bpf_filters *filters;
+	struct io_uring_bpf reg;
+	struct bpf_prog *prog;
+	struct sock_fprog fprog;
+	int ret;
+
+	if (copy_from_user(&reg, arg, sizeof(reg)))
+		return -EFAULT;
+	if (reg.cmd_type != IO_URING_BPF_CMD_FILTER)
+		return -EINVAL;
+	if (reg.cmd_flags || reg.resv)
+		return -EINVAL;
+
+	if (reg.filter.opcode >= IORING_OP_LAST)
+		return -EINVAL;
+	if (reg.filter.flags & ~IO_URING_BPF_FILTER_FLAGS)
+		return -EINVAL;
+	if (reg.filter.resv)
+		return -EINVAL;
+	if (!mem_is_zero(reg.filter.resv2, sizeof(reg.filter.resv2)))
+		return -EINVAL;
+	if (!reg.filter.filter_len || reg.filter.filter_len > BPF_MAXINSNS)
+		return -EINVAL;
+
+	fprog.len = reg.filter.filter_len;
+	fprog.filter = u64_to_user_ptr(reg.filter.filter_ptr);
+
+	ret = bpf_prog_create_from_user(&prog, &fprog,
+					io_uring_check_cbpf_filter, false);
+	if (ret)
+		return ret;
+
+	/*
+	 * No existing filters, allocate set.
+	 */
+	filters = res->bpf_filters;
+	if (!filters) {
+		filters = io_new_bpf_filters();
+		if (IS_ERR(filters)) {
+			ret = PTR_ERR(filters);
+			goto err_prog;
+		}
+	}
+
+	filter = kzalloc(sizeof(*filter), GFP_KERNEL_ACCOUNT);
+	if (!filter) {
+		ret = -ENOMEM;
+		goto err;
+	}
+	filter->prog = prog;
+	res->bpf_filters = filters;
+
+	/*
+	 * Insert filter - if the current opcode already has a filter
+	 * attached, add to the set.
+	 */
+	rcu_read_lock();
+	spin_lock_bh(&filters->lock);
+	old_filter = rcu_dereference(filters->filters[reg.filter.opcode]);
+	if (old_filter)
+		filter->next = old_filter;
+	rcu_assign_pointer(filters->filters[reg.filter.opcode], filter);
+
+	/*
+	 * If IO_URING_BPF_FILTER_DENY_REST is set, fill any unregistered
+	 * opcode with the dummy filter. That will cause them to be denied.
+	 */
+	if (reg.filter.flags & IO_URING_BPF_FILTER_DENY_REST) {
+		for (int i = 0; i < IORING_OP_LAST; i++) {
+			if (i == reg.filter.opcode)
+				continue;
+			old_filter = rcu_dereference(filters->filters[i]);
+			if (old_filter)
+				continue;
+			rcu_assign_pointer(filters->filters[i], &dummy_filter);
+		}
+	}
+
+	spin_unlock_bh(&filters->lock);
+	rcu_read_unlock();
+	return 0;
+err:
+	if (filters != res->bpf_filters)
+		__io_put_bpf_filters(filters);
+err_prog:
+	bpf_prog_destroy(prog);
+	return ret;
+}
diff --git a/io_uring/bpf_filter.h b/io_uring/bpf_filter.h
new file mode 100644
index 000000000000..27eae9705473
--- /dev/null
+++ b/io_uring/bpf_filter.h
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0
+#ifndef IO_URING_BPF_FILTER_H
+#define IO_URING_BPF_FILTER_H
+
+#include <uapi/linux/io_uring/bpf_filter.h>
+
+#ifdef CONFIG_IO_URING_BPF
+
+int __io_uring_run_bpf_filters(struct io_restriction *res, struct io_kiocb *req);
+
+int io_register_bpf_filter(struct io_restriction *res,
+			   struct io_uring_bpf __user *arg);
+
+void io_put_bpf_filters(struct io_restriction *res);
+
+static inline int io_uring_run_bpf_filters(struct io_restriction *res,
+					   struct io_kiocb *req)
+{
+	if (res->bpf_filters)
+		return __io_uring_run_bpf_filters(res, req);
+
+	return 0;
+}
+
+#else
+
+static inline int io_register_bpf_filter(struct io_restriction *res,
+					 struct io_uring_bpf __user *arg)
+{
+	return -EINVAL;
+}
+static inline int io_uring_run_bpf_filters(struct io_restriction *res,
+					   struct io_kiocb *req)
+{
+	return 0;
+}
+static inline void io_put_bpf_filters(struct io_restriction *res)
+{
+}
+#endif /* CONFIG_IO_URING_BPF */
+
+#endif
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index a50459238bee..9b9794dfc27a 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -94,6 +94,7 @@
 #include "alloc_cache.h"
 #include "eventfd.h"
 #include "wait.h"
+#include "bpf_filter.h"
 
 #define SQE_COMMON_FLAGS (IOSQE_FIXED_FILE | IOSQE_IO_LINK | \
 			  IOSQE_IO_HARDLINK | IOSQE_ASYNC)
@@ -1874,6 +1875,12 @@ static inline int io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
 	if (unlikely(ret))
 		return io_submit_fail_init(sqe, req, ret);
 
+	if (unlikely(ctx->restrictions.bpf_filters)) {
+		ret = io_uring_run_bpf_filters(&ctx->restrictions, req);
+		if (ret)
+			return io_submit_fail_init(sqe, req, ret);
+	}
+
 	trace_io_uring_submit_req(req);
 
 	/*
@@ -2161,6 +2168,7 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)
 	percpu_ref_exit(&ctx->refs);
 	free_uid(ctx->user);
 	io_req_caches_free(ctx);
+	io_put_bpf_filters(&ctx->restrictions);
 
 	WARN_ON_ONCE(ctx->nr_req_allocated);
 
diff --git a/io_uring/register.c b/io_uring/register.c
index 8551f13920dc..30957c2cb5eb 100644
--- a/io_uring/register.c
+++ b/io_uring/register.c
@@ -33,6 +33,7 @@
 #include "memmap.h"
 #include "zcrx.h"
 #include "query.h"
+#include "bpf_filter.h"
 
 #define IORING_MAX_RESTRICTIONS	(IORING_RESTRICTION_LAST + \
 				 IORING_REGISTER_LAST + IORING_OP_LAST)
@@ -830,6 +831,13 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 	case IORING_REGISTER_ZCRX_CTRL:
 		ret = io_zcrx_ctrl(ctx, arg, nr_args);
 		break;
+	case IORING_REGISTER_BPF_FILTER:
+		ret = -EINVAL;
+
+		if (nr_args != 1)
+			break;
+		ret = io_register_bpf_filter(&ctx->restrictions, arg);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
-- 
2.51.0

[PATCH 2/7] io_uring/net: allow filtering on IORING_OP_SOCKET data

[Jens Axboe]

Example population method for the BPF based opcode filtering. This
exposes the socket family, type, and protocol to a registered BPF
filter. This in turn enables the filter to make decisions based on
what was passed in to the IORING_OP_SOCKET request type.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/uapi/linux/io_uring/bpf_filter.h |  7 +++++++
 io_uring/bpf_filter.c                    | 11 +++++++++++
 io_uring/net.c                           |  9 +++++++++
 io_uring/net.h                           |  6 ++++++
 4 files changed, 33 insertions(+)

diff --git a/include/uapi/linux/io_uring/bpf_filter.h b/include/uapi/linux/io_uring/bpf_filter.h
index 2d4d0e5743e4..4dbc89bbbf10 100644
--- a/include/uapi/linux/io_uring/bpf_filter.h
+++ b/include/uapi/linux/io_uring/bpf_filter.h
@@ -16,6 +16,13 @@ struct io_uring_bpf_ctx {
 	__u8	sqe_flags;
 	__u8	pdu_size;	/* size of aux data for filter */
 	__u8	pad[5];
+	union {
+		struct {
+			__u32	family;
+			__u32	type;
+			__u32	protocol;
+		} socket;
+	};
 };
 
 enum {
diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
index 5207226d72ea..889fa915fa54 100644
--- a/io_uring/bpf_filter.c
+++ b/io_uring/bpf_filter.c
@@ -30,6 +30,17 @@ static void io_uring_populate_bpf_ctx(struct io_uring_bpf_ctx *bctx,
 	/* clear residual, anything from pdu_size and below */
 	memset((void *) bctx + offsetof(struct io_uring_bpf_ctx, pdu_size), 0,
 		sizeof(*bctx) - offsetof(struct io_uring_bpf_ctx, pdu_size));
+
+	/*
+	 * Opcodes can provide a handler fo populating more data into bctx,
+	 * for filters to use.
+	 */
+	switch (req->opcode) {
+	case IORING_OP_SOCKET:
+		bctx->pdu_size = sizeof(bctx->socket);
+		io_socket_bpf_populate(bctx, req);
+		break;
+	}
 }
 
 /*
diff --git a/io_uring/net.c b/io_uring/net.c
index 519ea055b761..4fcba36bd0bb 100644
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -1699,6 +1699,15 @@ int io_accept(struct io_kiocb *req, unsigned int issue_flags)
 	return IOU_COMPLETE;
 }
 
+void io_socket_bpf_populate(struct io_uring_bpf_ctx *bctx, struct io_kiocb *req)
+{
+	struct io_socket *sock = io_kiocb_to_cmd(req, struct io_socket);
+
+	bctx->socket.family = sock->domain;
+	bctx->socket.type = sock->type;
+	bctx->socket.protocol = sock->protocol;
+}
+
 int io_socket_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	struct io_socket *sock = io_kiocb_to_cmd(req, struct io_socket);
diff --git a/io_uring/net.h b/io_uring/net.h
index 43e5ce5416b7..a862960a3bb9 100644
--- a/io_uring/net.h
+++ b/io_uring/net.h
@@ -3,6 +3,7 @@
 #include <linux/net.h>
 #include <linux/uio.h>
 #include <linux/io_uring_types.h>
+#include <uapi/linux/io_uring/bpf_filter.h>
 
 struct io_async_msghdr {
 #if defined(CONFIG_NET)
@@ -44,6 +45,7 @@ int io_accept(struct io_kiocb *req, unsigned int issue_flags);
 
 int io_socket_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_socket(struct io_kiocb *req, unsigned int issue_flags);
+void io_socket_bpf_populate(struct io_uring_bpf_ctx *bctx, struct io_kiocb *req);
 
 int io_connect_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_connect(struct io_kiocb *req, unsigned int issue_flags);
@@ -64,4 +66,8 @@ void io_netmsg_cache_free(const void *entry);
 static inline void io_netmsg_cache_free(const void *entry)
 {
 }
+static inline void io_socket_bpf_populate(struct io_uring_bpf_ctx *bctx,
+					  struct io_kiocb *req)
+{
+}
 #endif
-- 
2.51.0

[PATCH 3/7] io_uring/bpf_filter: allow filtering on contents of struct open_how

[Jens Axboe]

This adds custom filtering for IORING_OP_OPENAT and IORING_OP_OPENAT2,
where the open_how flags, mode, and resolve can be checked by filters.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/uapi/linux/io_uring/bpf_filter.h | 5 +++++
 io_uring/bpf_filter.c                    | 6 ++++++
 io_uring/openclose.c                     | 9 +++++++++
 io_uring/openclose.h                     | 3 +++
 4 files changed, 23 insertions(+)

diff --git a/include/uapi/linux/io_uring/bpf_filter.h b/include/uapi/linux/io_uring/bpf_filter.h
index 4dbc89bbbf10..220351b81bc0 100644
--- a/include/uapi/linux/io_uring/bpf_filter.h
+++ b/include/uapi/linux/io_uring/bpf_filter.h
@@ -22,6 +22,11 @@ struct io_uring_bpf_ctx {
 			__u32	type;
 			__u32	protocol;
 		} socket;
+		struct {
+			__u64	flags;
+			__u64	mode;
+			__u64	resolve;
+		} open;
 	};
 };
 
diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
index 889fa915fa54..ff723ec44828 100644
--- a/io_uring/bpf_filter.c
+++ b/io_uring/bpf_filter.c
@@ -12,6 +12,7 @@
 #include "io_uring.h"
 #include "bpf_filter.h"
 #include "net.h"
+#include "openclose.h"
 
 struct io_bpf_filter {
 	struct bpf_prog		*prog;
@@ -40,6 +41,11 @@ static void io_uring_populate_bpf_ctx(struct io_uring_bpf_ctx *bctx,
 		bctx->pdu_size = sizeof(bctx->socket);
 		io_socket_bpf_populate(bctx, req);
 		break;
+	case IORING_OP_OPENAT:
+	case IORING_OP_OPENAT2:
+		bctx->pdu_size = sizeof(bctx->open);
+		io_openat_bpf_populate(bctx, req);
+		break;
 	}
 }
 
diff --git a/io_uring/openclose.c b/io_uring/openclose.c
index 15dde9bd6ff6..31c687adf873 100644
--- a/io_uring/openclose.c
+++ b/io_uring/openclose.c
@@ -85,6 +85,15 @@ static int __io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe
 	return 0;
 }
 
+void io_openat_bpf_populate(struct io_uring_bpf_ctx *bctx, struct io_kiocb *req)
+{
+	struct io_open *open = io_kiocb_to_cmd(req, struct io_open);
+
+	bctx->open.flags = open->how.flags;
+	bctx->open.mode = open->how.mode;
+	bctx->open.resolve = open->how.resolve;
+}
+
 int io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	struct io_open *open = io_kiocb_to_cmd(req, struct io_open);
diff --git a/io_uring/openclose.h b/io_uring/openclose.h
index 4ca2a9935abc..566739920658 100644
--- a/io_uring/openclose.h
+++ b/io_uring/openclose.h
@@ -1,11 +1,14 @@
 // SPDX-License-Identifier: GPL-2.0
 
+#include "bpf_filter.h"
+
 int __io_close_fixed(struct io_ring_ctx *ctx, unsigned int issue_flags,
 		     unsigned int offset);
 
 int io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_openat(struct io_kiocb *req, unsigned int issue_flags);
 void io_open_cleanup(struct io_kiocb *req);
+void io_openat_bpf_populate(struct io_uring_bpf_ctx *bctx, struct io_kiocb *req);
 
 int io_openat2_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_openat2(struct io_kiocb *req, unsigned int issue_flags);
-- 
2.51.0

[PATCH 4/7] io_uring/bpf_filter: cache lookup table in ctx->bpf_filters

[Jens Axboe]

Currently a few pointer dereferences need to be made to both check if
BPF filters are installed, and then also to retrieve the actual filter
for the opcode. Cache the table in ctx->bpf_filters to avoid that.

Add a bit of debug info on ring exit to show if we ever got this wrong.
Small risk of that given that the table is currently only updated in one
spot, but once task forking is enabled, that will add one more spot.

Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/io_uring_types.h |  2 ++
 io_uring/bpf_filter.c          |  7 ++++---
 io_uring/bpf_filter.h          | 10 +++++-----
 io_uring/io_uring.c            | 11 +++++++++--
 io_uring/register.c            |  3 +++
 5 files changed, 23 insertions(+), 10 deletions(-)

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 74bf98362876..7617df247238 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -287,6 +287,8 @@ struct io_ring_ctx {
 
 		struct task_struct	*submitter_task;
 		struct io_rings		*rings;
+		/* cache of ->restrictions.bpf_filters->filters */
+		struct io_bpf_filter __rcu	**bpf_filters;
 		struct percpu_ref	refs;
 
 		clockid_t		clockid;
diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
index ff723ec44828..1409d625b686 100644
--- a/io_uring/bpf_filter.c
+++ b/io_uring/bpf_filter.c
@@ -58,14 +58,15 @@ static void io_uring_populate_bpf_ctx(struct io_uring_bpf_ctx *bctx,
  * __io_uring_run_bpf_filters() returns 0 on success, allow running the
  * request, and -EACCES when a request is denied.
  */
-int __io_uring_run_bpf_filters(struct io_restriction *res, struct io_kiocb *req)
+int __io_uring_run_bpf_filters(struct io_bpf_filter __rcu **filters,
+			       struct io_kiocb *req)
 {
 	struct io_bpf_filter *filter;
 	struct io_uring_bpf_ctx bpf_ctx;
 	int ret;
 
 	/* Fast check for existence of filters outside of RCU */
-	if (!rcu_access_pointer(res->bpf_filters->filters[req->opcode]))
+	if (!rcu_access_pointer(filters[req->opcode]))
 		return 0;
 
 	/*
@@ -73,7 +74,7 @@ int __io_uring_run_bpf_filters(struct io_restriction *res, struct io_kiocb *req)
 	 * of what we expect, io_init_req() does this.
 	 */
 	guard(rcu)();
-	filter = rcu_dereference(res->bpf_filters->filters[req->opcode]);
+	filter = rcu_dereference(filters[req->opcode]);
 	if (!filter)
 		return 0;
 	else if (filter == &dummy_filter)
diff --git a/io_uring/bpf_filter.h b/io_uring/bpf_filter.h
index 27eae9705473..9f3cdb92eb16 100644
--- a/io_uring/bpf_filter.h
+++ b/io_uring/bpf_filter.h
@@ -6,18 +6,18 @@
 
 #ifdef CONFIG_IO_URING_BPF
 
-int __io_uring_run_bpf_filters(struct io_restriction *res, struct io_kiocb *req);
+int __io_uring_run_bpf_filters(struct io_bpf_filter __rcu **filters, struct io_kiocb *req);
 
 int io_register_bpf_filter(struct io_restriction *res,
 			   struct io_uring_bpf __user *arg);
 
 void io_put_bpf_filters(struct io_restriction *res);
 
-static inline int io_uring_run_bpf_filters(struct io_restriction *res,
+static inline int io_uring_run_bpf_filters(struct io_bpf_filter __rcu **filters,
 					   struct io_kiocb *req)
 {
-	if (res->bpf_filters)
-		return __io_uring_run_bpf_filters(res, req);
+	if (filters)
+		return __io_uring_run_bpf_filters(filters, req);
 
 	return 0;
 }
@@ -29,7 +29,7 @@ static inline int io_register_bpf_filter(struct io_restriction *res,
 {
 	return -EINVAL;
 }
-static inline int io_uring_run_bpf_filters(struct io_restriction *res,
+static inline int io_uring_run_bpf_filters(struct io_bpf_filter __rcu **filters,
 					   struct io_kiocb *req)
 {
 	return 0;
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 9b9794dfc27a..049454278563 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1875,8 +1875,8 @@ static inline int io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
 	if (unlikely(ret))
 		return io_submit_fail_init(sqe, req, ret);
 
-	if (unlikely(ctx->restrictions.bpf_filters)) {
-		ret = io_uring_run_bpf_filters(&ctx->restrictions, req);
+	if (unlikely(ctx->bpf_filters)) {
+		ret = io_uring_run_bpf_filters(ctx->bpf_filters, req);
 		if (ret)
 			return io_submit_fail_init(sqe, req, ret);
 	}
@@ -2168,6 +2168,13 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)
 	percpu_ref_exit(&ctx->refs);
 	free_uid(ctx->user);
 	io_req_caches_free(ctx);
+
+	if (ctx->restrictions.bpf_filters) {
+		WARN_ON_ONCE(ctx->bpf_filters !=
+			     ctx->restrictions.bpf_filters->filters);
+	} else {
+		WARN_ON_ONCE(ctx->bpf_filters);
+	}
 	io_put_bpf_filters(&ctx->restrictions);
 
 	WARN_ON_ONCE(ctx->nr_req_allocated);
diff --git a/io_uring/register.c b/io_uring/register.c
index 30957c2cb5eb..40de9b8924b9 100644
--- a/io_uring/register.c
+++ b/io_uring/register.c
@@ -837,6 +837,9 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 		if (nr_args != 1)
 			break;
 		ret = io_register_bpf_filter(&ctx->restrictions, arg);
+		if (!ret)
+			WRITE_ONCE(ctx->bpf_filters,
+				   ctx->restrictions.bpf_filters->filters);
 		break;
 	default:
 		ret = -EINVAL;
-- 
2.51.0

[PATCH 5/7] io_uring/bpf_filter: add ref counts to struct io_bpf_filter

[Jens Axboe]

In preparation for allowing inheritance of BPF filters and filter
tables, add a reference count to the filter. This allows multiple tables
to safely include the same filter.

Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/bpf_filter.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
index 1409d625b686..b94944ab8442 100644
--- a/io_uring/bpf_filter.c
+++ b/io_uring/bpf_filter.c
@@ -15,6 +15,7 @@
 #include "openclose.h"
 
 struct io_bpf_filter {
+	refcount_t		refs;
 	struct bpf_prog		*prog;
 	struct io_bpf_filter	*next;
 };
@@ -125,6 +126,11 @@ static void io_free_bpf_filters(struct rcu_head *head)
 			 */
 			if (f == &dummy_filter)
 				break;
+
+			/* Someone still holds a ref, stop iterating. */
+			if (!refcount_dec_and_test(&f->refs))
+				break;
+
 			bpf_prog_destroy(f->prog);
 			kfree(f);
 			f = next;
@@ -298,6 +304,7 @@ int io_register_bpf_filter(struct io_restriction *res,
 		ret = -ENOMEM;
 		goto err;
 	}
+	refcount_set(&filter->refs, 1);
 	filter->prog = prog;
 	res->bpf_filters = filters;
 
-- 
2.51.0

[PATCH 6/7] io_uring: add task fork hook

[Jens Axboe]

Called when copy_process() is called to copy state to a new child.
Right now this is just a stub, but will be used shortly to properly
handle fork'ing of task based io_uring restrictions.

Reviewed-by: Christian Brauner (Microsoft) <brauner@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/io_uring.h | 14 +++++++++++++-
 include/linux/sched.h    |  1 +
 io_uring/tctx.c          | 25 ++++++++++++++++---------
 kernel/fork.c            |  5 +++++
 4 files changed, 35 insertions(+), 10 deletions(-)

diff --git a/include/linux/io_uring.h b/include/linux/io_uring.h
index 85fe4e6b275c..d1aa4edfc2a5 100644
--- a/include/linux/io_uring.h
+++ b/include/linux/io_uring.h
@@ -12,6 +12,7 @@ void __io_uring_free(struct task_struct *tsk);
 void io_uring_unreg_ringfd(void);
 const char *io_uring_get_opcode(u8 opcode);
 bool io_is_uring_fops(struct file *file);
+int __io_uring_fork(struct task_struct *tsk);
 
 static inline void io_uring_files_cancel(void)
 {
@@ -25,9 +26,16 @@ static inline void io_uring_task_cancel(void)
 }
 static inline void io_uring_free(struct task_struct *tsk)
 {
-	if (tsk->io_uring)
+	if (tsk->io_uring || tsk->io_uring_restrict)
 		__io_uring_free(tsk);
 }
+static inline int io_uring_fork(struct task_struct *tsk)
+{
+	if (tsk->io_uring_restrict)
+		return __io_uring_fork(tsk);
+
+	return 0;
+}
 #else
 static inline void io_uring_task_cancel(void)
 {
@@ -46,6 +54,10 @@ static inline bool io_is_uring_fops(struct file *file)
 {
 	return false;
 }
+static inline int io_uring_fork(struct task_struct *tsk)
+{
+	return 0;
+}
 #endif
 
 #endif
diff --git a/include/linux/sched.h b/include/linux/sched.h
index d395f2810fac..9abbd11bb87c 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1190,6 +1190,7 @@ struct task_struct {
 
 #ifdef CONFIG_IO_URING
 	struct io_uring_task		*io_uring;
+	struct io_restriction		*io_uring_restrict;
 #endif
 
 	/* Namespaces: */
diff --git a/io_uring/tctx.c b/io_uring/tctx.c
index 5b66755579c0..d4f7698805e4 100644
--- a/io_uring/tctx.c
+++ b/io_uring/tctx.c
@@ -54,16 +54,18 @@ void __io_uring_free(struct task_struct *tsk)
 	 * node is stored in the xarray. Until that gets sorted out, attempt
 	 * an iteration here and warn if any entries are found.
 	 */
-	xa_for_each(&tctx->xa, index, node) {
-		WARN_ON_ONCE(1);
-		break;
-	}
-	WARN_ON_ONCE(tctx->io_wq);
-	WARN_ON_ONCE(tctx->cached_refs);
+	if (tctx) {
+		xa_for_each(&tctx->xa, index, node) {
+			WARN_ON_ONCE(1);
+			break;
+		}
+		WARN_ON_ONCE(tctx->io_wq);
+		WARN_ON_ONCE(tctx->cached_refs);
 
-	percpu_counter_destroy(&tctx->inflight);
-	kfree(tctx);
-	tsk->io_uring = NULL;
+		percpu_counter_destroy(&tctx->inflight);
+		kfree(tctx);
+		tsk->io_uring = NULL;
+	}
 }
 
 __cold int io_uring_alloc_task_context(struct task_struct *task,
@@ -351,3 +353,8 @@ int io_ringfd_unregister(struct io_ring_ctx *ctx, void __user *__arg,
 
 	return i ? i : ret;
 }
+
+int __io_uring_fork(struct task_struct *tsk)
+{
+	return 0;
+}
diff --git a/kernel/fork.c b/kernel/fork.c
index b1f3915d5f8e..08a2515380ec 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -97,6 +97,7 @@
 #include <linux/kasan.h>
 #include <linux/scs.h>
 #include <linux/io_uring.h>
+#include <linux/io_uring_types.h>
 #include <linux/bpf.h>
 #include <linux/stackprotector.h>
 #include <linux/user_events.h>
@@ -2129,6 +2130,9 @@ __latent_entropy struct task_struct *copy_process(
 
 #ifdef CONFIG_IO_URING
 	p->io_uring = NULL;
+	retval = io_uring_fork(p);
+	if (unlikely(retval))
+		goto bad_fork_cleanup_delayacct;
 #endif
 
 	p->default_timer_slack_ns = current->timer_slack_ns;
@@ -2525,6 +2529,7 @@ __latent_entropy struct task_struct *copy_process(
 	mpol_put(p->mempolicy);
 #endif
 bad_fork_cleanup_delayacct:
+	io_uring_free(p);
 	delayacct_tsk_free(p);
 bad_fork_cleanup_count:
 	dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1);
-- 
2.51.0

[PATCH 7/7] io_uring: allow registration of per-task restrictions

[Jens Axboe]

Currently io_uring supports restricting operations on a per-ring basis.
To use those, the ring must be setup in a disabled state by setting
IORING_SETUP_R_DISABLED. Then restrictions can be set for the ring, and
the ring can then be enabled.

This commit adds support for IORING_REGISTER_RESTRICTIONS with ring_fd
== -1, like the other "blind" register opcodes which work on the task
rather than a specific ring. This allows registration of the same kind
of restrictions as can been done on a specific ring, but with the task
itself. Once done, any ring created will inherit these restrictions.

If a restriction filter is registered with a task, then it's inherited
on fork for its children. Children may only further restrict operations,
not extend them.

Inheriting restrictions include both the classic
IORING_REGISTER_RESTRICTIONS based restrictions, as well as the BPF
filters that have been registered with the task via
IORING_REGISTER_BPF_FILTER.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/io_uring_types.h |  2 +
 include/uapi/linux/io_uring.h  |  7 +++
 io_uring/bpf_filter.c          | 86 +++++++++++++++++++++++++++++++++-
 io_uring/bpf_filter.h          |  6 +++
 io_uring/io_uring.c            | 33 +++++++++++++
 io_uring/io_uring.h            |  1 +
 io_uring/register.c            | 80 +++++++++++++++++++++++++++++++
 io_uring/tctx.c                | 17 +++++++
 8 files changed, 231 insertions(+), 1 deletion(-)

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 7617df247238..510d801b9a55 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -231,6 +231,8 @@ struct io_restriction {
 	DECLARE_BITMAP(register_op, IORING_REGISTER_LAST);
 	DECLARE_BITMAP(sqe_op, IORING_OP_LAST);
 	struct io_bpf_filters *bpf_filters;
+	/* ->bpf_filters needs COW on modification */
+	bool bpf_filters_cow;
 	u8 sqe_flags_allowed;
 	u8 sqe_flags_required;
 	/* IORING_OP_* restrictions exist */
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 94669b77fee8..aeeffcf27fee 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -808,6 +808,13 @@ struct io_uring_restriction {
 	__u32 resv2[3];
 };
 
+struct io_uring_task_restriction {
+	__u16 flags;
+	__u16 nr_res;
+	__u32 resv[3];
+	__DECLARE_FLEX_ARRAY(struct io_uring_restriction, restrictions);
+};
+
 struct io_uring_clock_register {
 	__u32	clockid;
 	__u32	__resv[3];
diff --git a/io_uring/bpf_filter.c b/io_uring/bpf_filter.c
index b94944ab8442..3816883a45ed 100644
--- a/io_uring/bpf_filter.c
+++ b/io_uring/bpf_filter.c
@@ -249,13 +249,77 @@ static int io_uring_check_cbpf_filter(struct sock_filter *filter,
 	return 0;
 }
 
+void io_bpf_filter_clone(struct io_restriction *dst, struct io_restriction *src)
+{
+	if (!src->bpf_filters)
+		return;
+
+	rcu_read_lock();
+	/*
+	 * If the src filter is going away, just ignore it.
+	 */
+	if (refcount_inc_not_zero(&src->bpf_filters->refs)) {
+		dst->bpf_filters = src->bpf_filters;
+		dst->bpf_filters_cow = true;
+	}
+	rcu_read_unlock();
+}
+
+/*
+ * Allocate a new struct io_bpf_filters. Used when a filter is cloned and
+ * modifications need to be made.
+ */
+static struct io_bpf_filters *io_bpf_filter_cow(struct io_restriction *src)
+{
+	struct io_bpf_filters *filters;
+	struct io_bpf_filter *srcf;
+	int i;
+
+	filters = io_new_bpf_filters();
+	if (IS_ERR(filters))
+		return filters;
+
+	/*
+	 * Iterate filters from src and assign in destination. Grabbing
+	 * a reference is enough, we don't need to duplicate the memory.
+	 * This is safe because filters are only ever appended to the
+	 * front of the list, hence the only memory ever touched inside
+	 * a filter is the refcount.
+	 */
+	rcu_read_lock();
+	for (i = 0; i < IORING_OP_LAST; i++) {
+		srcf = rcu_dereference(src->bpf_filters->filters[i]);
+		if (!srcf) {
+			continue;
+		} else if (srcf == &dummy_filter) {
+			rcu_assign_pointer(filters->filters[i], &dummy_filter);
+			continue;
+		}
+
+		/*
+		 * Getting a ref on the first node is enough, putting the
+		 * filter and iterating nodes to free will stop on the first
+		 * one that doesn't hit zero when dropping.
+		 */
+		if (!refcount_inc_not_zero(&srcf->refs))
+			goto err;
+		rcu_assign_pointer(filters->filters[i], srcf);
+	}
+	rcu_read_unlock();
+	return filters;
+err:
+	rcu_read_unlock();
+	__io_put_bpf_filters(filters);
+	return ERR_PTR(-EBUSY);
+}
+
 #define IO_URING_BPF_FILTER_FLAGS	IO_URING_BPF_FILTER_DENY_REST
 
 int io_register_bpf_filter(struct io_restriction *res,
 			   struct io_uring_bpf __user *arg)
 {
+	struct io_bpf_filters *filters, *old_filters = NULL;
 	struct io_bpf_filter *filter, *old_filter;
-	struct io_bpf_filters *filters;
 	struct io_uring_bpf reg;
 	struct bpf_prog *prog;
 	struct sock_fprog fprog;
@@ -297,6 +361,17 @@ int io_register_bpf_filter(struct io_restriction *res,
 			ret = PTR_ERR(filters);
 			goto err_prog;
 		}
+	} else if (res->bpf_filters_cow) {
+		filters = io_bpf_filter_cow(res);
+		if (IS_ERR(filters)) {
+			ret = PTR_ERR(filters);
+			goto err_prog;
+		}
+		/*
+		 * Stash old filters, we'll put them once we know we'll
+		 * succeed. Until then, res->bpf_filters is left untouched.
+		 */
+		old_filters = res->bpf_filters;
 	}
 
 	filter = kzalloc(sizeof(*filter), GFP_KERNEL_ACCOUNT);
@@ -306,6 +381,15 @@ int io_register_bpf_filter(struct io_restriction *res,
 	}
 	refcount_set(&filter->refs, 1);
 	filter->prog = prog;
+
+	/*
+	 * Success - install the new filter set now. If we did COW, put
+	 * the old filters as we're replacing them.
+	 */
+	if (old_filters) {
+		__io_put_bpf_filters(old_filters);
+		res->bpf_filters_cow = false;
+	}
 	res->bpf_filters = filters;
 
 	/*
diff --git a/io_uring/bpf_filter.h b/io_uring/bpf_filter.h
index 9f3cdb92eb16..66a776cf25b4 100644
--- a/io_uring/bpf_filter.h
+++ b/io_uring/bpf_filter.h
@@ -13,6 +13,8 @@ int io_register_bpf_filter(struct io_restriction *res,
 
 void io_put_bpf_filters(struct io_restriction *res);
 
+void io_bpf_filter_clone(struct io_restriction *dst, struct io_restriction *src);
+
 static inline int io_uring_run_bpf_filters(struct io_bpf_filter __rcu **filters,
 					   struct io_kiocb *req)
 {
@@ -37,6 +39,10 @@ static inline int io_uring_run_bpf_filters(struct io_bpf_filter __rcu **filters,
 static inline void io_put_bpf_filters(struct io_restriction *res)
 {
 }
+static inline void io_bpf_filter_clone(struct io_restriction *dst,
+				       struct io_restriction *src)
+{
+}
 #endif /* CONFIG_IO_URING_BPF */
 
 #endif
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 049454278563..e43c5283b23a 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -2880,6 +2880,32 @@ int io_prepare_config(struct io_ctx_config *config)
 	return 0;
 }
 
+void io_restriction_clone(struct io_restriction *dst, struct io_restriction *src)
+{
+	memcpy(&dst->register_op, &src->register_op, sizeof(dst->register_op));
+	memcpy(&dst->sqe_op, &src->sqe_op, sizeof(dst->sqe_op));
+	dst->sqe_flags_allowed = src->sqe_flags_allowed;
+	dst->sqe_flags_required = src->sqe_flags_required;
+	dst->op_registered = src->op_registered;
+	dst->reg_registered = src->reg_registered;
+
+	io_bpf_filter_clone(dst, src);
+}
+
+static void io_ctx_restriction_clone(struct io_ring_ctx *ctx,
+				     struct io_restriction *src)
+{
+	struct io_restriction *dst = &ctx->restrictions;
+
+	io_restriction_clone(dst, src);
+	if (dst->bpf_filters)
+		WRITE_ONCE(ctx->bpf_filters, dst->bpf_filters->filters);
+	if (dst->op_registered)
+		ctx->op_restricted = 1;
+	if (dst->reg_registered)
+		ctx->reg_restricted = 1;
+}
+
 static __cold int io_uring_create(struct io_ctx_config *config)
 {
 	struct io_uring_params *p = &config->p;
@@ -2940,6 +2966,13 @@ static __cold int io_uring_create(struct io_ctx_config *config)
 	else
 		ctx->notify_method = TWA_SIGNAL;
 
+	/*
+	 * If the current task has restrictions enabled, then copy them to
+	 * our newly created ring and mark it as registered.
+	 */
+	if (current->io_uring_restrict)
+		io_ctx_restriction_clone(ctx, current->io_uring_restrict);
+
 	/*
 	 * This is just grabbed for accounting purposes. When a process exits,
 	 * the mm is exited and dropped before the files, hence we need to hang
diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h
index 29b8f90fdabf..a08d78c716f8 100644
--- a/io_uring/io_uring.h
+++ b/io_uring/io_uring.h
@@ -197,6 +197,7 @@ void io_task_refs_refill(struct io_uring_task *tctx);
 bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
 
 void io_activate_pollwq(struct io_ring_ctx *ctx);
+void io_restriction_clone(struct io_restriction *dst, struct io_restriction *src);
 
 static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
 {
diff --git a/io_uring/register.c b/io_uring/register.c
index 40de9b8924b9..af4815bc11d6 100644
--- a/io_uring/register.c
+++ b/io_uring/register.c
@@ -190,6 +190,82 @@ static __cold int io_register_restrictions(struct io_ring_ctx *ctx,
 	return 0;
 }
 
+static int io_register_restrictions_task(void __user *arg, unsigned int nr_args)
+{
+	struct io_uring_task_restriction __user *ures = arg;
+	struct io_uring_task_restriction tres;
+	struct io_restriction *res;
+	int ret;
+
+	/* Disallow if task already has registered restrictions */
+	if (current->io_uring_restrict)
+		return -EPERM;
+	/*
+	 * Similar to seccomp, disallow setting a filter if task_no_new_privs
+	 * is true and we're not CAP_SYS_ADMIN.
+	 */
+	if (!task_no_new_privs(current) &&
+	    !ns_capable_noaudit(current_user_ns(), CAP_SYS_ADMIN))
+		return -EACCES;
+	if (nr_args != 1)
+		return -EINVAL;
+
+	if (copy_from_user(&tres, arg, sizeof(tres)))
+		return -EFAULT;
+
+	if (tres.flags)
+		return -EINVAL;
+	if (!mem_is_zero(tres.resv, sizeof(tres.resv)))
+		return -EINVAL;
+
+	res = kzalloc(sizeof(*res), GFP_KERNEL_ACCOUNT);
+	if (!res)
+		return -ENOMEM;
+
+	ret = io_parse_restrictions(ures->restrictions, tres.nr_res, res);
+	if (ret < 0) {
+		kfree(res);
+		return ret;
+	}
+	current->io_uring_restrict = res;
+	return 0;
+}
+
+static int io_register_bpf_filter_task(void __user *arg, unsigned int nr_args)
+{
+	struct io_restriction *res;
+	int ret;
+
+	/*
+	 * Similar to seccomp, disallow setting a filter if task_no_new_privs
+	 * is true and we're not CAP_SYS_ADMIN.
+	 */
+	if (!task_no_new_privs(current) &&
+	    !ns_capable_noaudit(current_user_ns(), CAP_SYS_ADMIN))
+		return -EACCES;
+
+	if (nr_args != 1)
+		return -EINVAL;
+
+	/* If no task restrictions exist, setup a new set */
+	res = current->io_uring_restrict;
+	if (!res) {
+		res = kzalloc(sizeof(*res), GFP_KERNEL_ACCOUNT);
+		if (!res)
+			return -ENOMEM;
+	}
+
+	ret = io_register_bpf_filter(res, arg);
+	if (ret) {
+		if (res != current->io_uring_restrict)
+			kfree(res);
+		return ret;
+	}
+	if (!current->io_uring_restrict)
+		current->io_uring_restrict = res;
+	return 0;
+}
+
 static int io_register_enable_rings(struct io_ring_ctx *ctx)
 {
 	if (!(ctx->flags & IORING_SETUP_R_DISABLED))
@@ -912,6 +988,10 @@ static int io_uring_register_blind(unsigned int opcode, void __user *arg,
 		return io_uring_register_send_msg_ring(arg, nr_args);
 	case IORING_REGISTER_QUERY:
 		return io_query(arg, nr_args);
+	case IORING_REGISTER_RESTRICTIONS:
+		return io_register_restrictions_task(arg, nr_args);
+	case IORING_REGISTER_BPF_FILTER:
+		return io_register_bpf_filter_task(arg, nr_args);
 	}
 	return -EINVAL;
 }
diff --git a/io_uring/tctx.c b/io_uring/tctx.c
index d4f7698805e4..e3da31fdf16f 100644
--- a/io_uring/tctx.c
+++ b/io_uring/tctx.c
@@ -11,6 +11,7 @@
 
 #include "io_uring.h"
 #include "tctx.h"
+#include "bpf_filter.h"
 
 static struct io_wq *io_init_wq_offload(struct io_ring_ctx *ctx,
 					struct task_struct *task)
@@ -66,6 +67,11 @@ void __io_uring_free(struct task_struct *tsk)
 		kfree(tctx);
 		tsk->io_uring = NULL;
 	}
+	if (tsk->io_uring_restrict) {
+		io_put_bpf_filters(tsk->io_uring_restrict);
+		kfree(tsk->io_uring_restrict);
+		tsk->io_uring_restrict = NULL;
+	}
 }
 
 __cold int io_uring_alloc_task_context(struct task_struct *task,
@@ -356,5 +362,16 @@ int io_ringfd_unregister(struct io_ring_ctx *ctx, void __user *__arg,
 
 int __io_uring_fork(struct task_struct *tsk)
 {
+	struct io_restriction *res, *src = tsk->io_uring_restrict;
+
+	/* Don't leave it dangling on error */
+	tsk->io_uring_restrict = NULL;
+
+	res = kzalloc(sizeof(*res), GFP_KERNEL_ACCOUNT);
+	if (!res)
+		return -ENOMEM;
+
+	tsk->io_uring_restrict = res;
+	io_restriction_clone(res, src);
 	return 0;
 }
-- 
2.51.0