Merge tag 'io_uring-7.3-20261002' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linuxgrafted

Pull io_uring fixes from Jens Axboe:

 - Fix a task_work add use-after-free with SQPOLL.

   The sqpoll thread could pop and complete the last request while
   io_req_normal_work_add() was still looking at them after the mpscq
   push.

   Use the same approach as DEFER_TASKRUN to protect from that, holding
   an RCU read lock across the add, and have exit wait for an RCU grace
   period for SQPOLL rings as well.

 - CQE32 ring fixes: correct the free entry check for 32b CQEs, zero the
   big_cqe for aux CQEs, and only post the dummy skip CQE on CQE_MIXED
   rings

 - Mark the source filter table as COW when cloning bpf filters, so
   registering another filter on the source doesn't modify the shared
   table in place

 - Initialize the task context before running the BPF loop

 - Requeue zcrx multishot receives stopped by a local resource

 - End a TX_TIMESTAMP multishot cmd when the CQ is full (lollipopkit)

* tag 'io_uring-7.3-20261002' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux:
  io_uring: fix task_work add use-after-free with SQPOLL
  io_uring/cmd_net: end TX_TIMESTAMP multishot when the CQ is full
  io_uring/zcrx: requeue multishot receives stopped by a local resource
  io_uring: initialize task context before running the BPF loop
  io_uring: zero big_cqe for aux CQEs on CQE32 rings
  io_uring: fix free entry check for 32b CQEs on CQE32 rings
  io_uring: only post the dummy skip CQE on CQE_MIXED rings
  io_uring/bpf_filter: mark source as COW when cloning filters

1 files changed, 121 insertions, 0 deletions

diff --git a/drivers/misc/sram-exec.c b/drivers/misc/sram-exec.c
new file mode 100644
index 000000000..b71dbbd73
--- /dev/null
+++ b/drivers/misc/sram-exec.c
@@ -0,0 +1,121 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * SRAM protect-exec region helper functions
+ *
+ * Copyright (C) 2017 Texas Instruments Incorporated - https://www.ti.com/
+ *	Dave Gerlach
+ */
+
+#include <linux/device.h>
+#include <linux/genalloc.h>
+#include <linux/mm.h>
+#include <linux/sram.h>
+#include <linux/set_memory.h>
+
+#include <asm/fncpy.h>
+
+#include "sram.h"
+
+static DEFINE_MUTEX(exec_pool_list_mutex);
+static LIST_HEAD(exec_pool_list);
+
+int sram_check_protect_exec(struct sram_dev *sram, struct sram_reserve *block,
+			    struct sram_partition *part)
+{
+	unsigned long base = (unsigned long)part->base;
+	unsigned long end = base + block->size;
+
+	if (!PAGE_ALIGNED(base) || !PAGE_ALIGNED(end)) {
+		dev_err(sram->dev,
+			"SRAM pool marked with 'protect-exec' is not page aligned and will not be created.\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+int sram_add_protect_exec(struct sram_partition *part)
+{
+	mutex_lock(&exec_pool_list_mutex);
+	list_add_tail(&part->list, &exec_pool_list);
+	mutex_unlock(&exec_pool_list_mutex);
+
+	return 0;
+}
+
+/**
+ * sram_exec_copy - copy data to a protected executable region of sram
+ *
+ * @pool: struct gen_pool retrieved that is part of this sram
+ * @dst: Destination address for the copy, that must be inside pool
+ * @src: Source address for the data to copy
+ * @size: Size of copy to perform, which starting from dst, must reside in pool
+ *
+ * Return: Address for copied data that can safely be called through function
+ *	   pointer, or NULL if problem.
+ *
+ * This helper function allows sram driver to act as central control location
+ * of 'protect-exec' pools which are normal sram pools but are always set
+ * read-only and executable except when copying data to them, at which point
+ * they are set to read-write non-executable, to make sure no memory is
+ * writeable and executable at the same time. This region must be page-aligned
+ * and is checked during probe, otherwise page attribute manipulation would
+ * not be possible. Care must be taken to only call the returned address as
+ * dst address is not guaranteed to be safely callable.
+ *
+ * NOTE: This function uses the fncpy macro to move code to the executable
+ * region. Some architectures have strict requirements for relocating
+ * executable code, so fncpy is a macro that must be defined by any arch
+ * making use of this functionality that guarantees a safe copy of exec
+ * data and returns a safe address that can be called as a C function
+ * pointer.
+ */
+void *sram_exec_copy(struct gen_pool *pool, void *dst, void *src,
+		     size_t size)
+{
+	struct sram_partition *part = NULL, *p;
+	unsigned long base;
+	int pages;
+	void *dst_cpy;
+	int ret;
+
+	mutex_lock(&exec_pool_list_mutex);
+	list_for_each_entry(p, &exec_pool_list, list) {
+		if (p->pool == pool)
+			part = p;
+	}
+	mutex_unlock(&exec_pool_list_mutex);
+
+	if (!part)
+		return NULL;
+
+	if (!gen_pool_has_addr(pool, (unsigned long)dst, size))
+		return NULL;
+
+	base = (unsigned long)part->base;
+	pages = PAGE_ALIGN(size) / PAGE_SIZE;
+
+	mutex_lock(&part->lock);
+
+	ret = set_memory_nx((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+	ret = set_memory_rw((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+
+	dst_cpy = fncpy(dst, src, size);
+
+	ret = set_memory_rox((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+
+	mutex_unlock(&part->lock);
+
+	return dst_cpy;
+
+error_out:
+	mutex_unlock(&part->lock);
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(sram_exec_copy);