zephyr/subsys/rtio/rtio_handlers.c

136 lines
3.7 KiB
C

/*
* Copyright (c) 2022 Intel Corporation
*
* SPDX-License-Identifier: Apache-2.0
*/
#include <zephyr/kernel.h>
#include <stdbool.h>
#include <zephyr/rtio/rtio.h>
#include <zephyr/internal/syscall_handler.h>
/**
* Verify each SQE type operation and its fields ensuring
* the iodev is a valid accessible k_object (if given) and
* the buffer pointers are valid accessible memory by the calling
* thread.
*
* Each op code that is acceptable from user mode must also be validated.
*/
static inline bool rtio_vrfy_sqe(struct rtio_sqe *sqe)
{
if (sqe->iodev != NULL && K_SYSCALL_OBJ(sqe->iodev, K_OBJ_RTIO_IODEV)) {
return false;
}
bool valid_sqe = true;
switch (sqe->op) {
case RTIO_OP_NOP:
break;
case RTIO_OP_TX:
valid_sqe &= K_SYSCALL_MEMORY(sqe->tx.buf, sqe->tx.buf_len, false);
break;
case RTIO_OP_RX:
if ((sqe->flags & RTIO_SQE_MEMPOOL_BUFFER) == 0) {
valid_sqe &= K_SYSCALL_MEMORY(sqe->rx.buf, sqe->rx.buf_len, true);
}
break;
case RTIO_OP_TINY_TX:
break;
case RTIO_OP_TXRX:
valid_sqe &= K_SYSCALL_MEMORY(sqe->txrx.tx_buf, sqe->txrx.buf_len, true);
valid_sqe &= K_SYSCALL_MEMORY(sqe->txrx.rx_buf, sqe->txrx.buf_len, true);
break;
default:
/* RTIO OP must be known and allowable from user mode
* otherwise it is invalid
*/
valid_sqe = false;
}
return valid_sqe;
}
static inline void z_vrfy_rtio_release_buffer(struct rtio *r, void *buff, uint32_t buff_len)
{
K_OOPS(K_SYSCALL_OBJ(r, K_OBJ_RTIO));
z_impl_rtio_release_buffer(r, buff, buff_len);
}
#include <zephyr/syscalls/rtio_release_buffer_mrsh.c>
static inline int z_vrfy_rtio_cqe_get_mempool_buffer(const struct rtio *r, struct rtio_cqe *cqe,
uint8_t **buff, uint32_t *buff_len)
{
K_OOPS(K_SYSCALL_OBJ(r, K_OBJ_RTIO));
K_OOPS(K_SYSCALL_MEMORY_READ(cqe, sizeof(struct rtio_cqe)));
K_OOPS(K_SYSCALL_MEMORY_READ(buff, sizeof(void *)));
K_OOPS(K_SYSCALL_MEMORY_READ(buff_len, sizeof(uint32_t)));
return z_impl_rtio_cqe_get_mempool_buffer(r, cqe, buff, buff_len);
}
#include <zephyr/syscalls/rtio_cqe_get_mempool_buffer_mrsh.c>
static inline int z_vrfy_rtio_sqe_cancel(struct rtio_sqe *sqe)
{
return z_impl_rtio_sqe_cancel(sqe);
}
#include <zephyr/syscalls/rtio_sqe_cancel_mrsh.c>
static inline int z_vrfy_rtio_sqe_copy_in_get_handles(struct rtio *r, const struct rtio_sqe *sqes,
struct rtio_sqe **handle, size_t sqe_count)
{
K_OOPS(K_SYSCALL_OBJ(r, K_OBJ_RTIO));
K_OOPS(K_SYSCALL_MEMORY_ARRAY_READ(sqes, sqe_count, sizeof(struct rtio_sqe)));
struct rtio_sqe *sqe;
uint32_t acquirable = rtio_sqe_acquirable(r);
if (acquirable < sqe_count) {
return -ENOMEM;
}
for (int i = 0; i < sqe_count; i++) {
sqe = rtio_sqe_acquire(r);
__ASSERT_NO_MSG(sqe != NULL);
if (handle != NULL && i == 0) {
*handle = sqe;
}
*sqe = sqes[i];
if (!rtio_vrfy_sqe(sqe)) {
rtio_sqe_drop_all(r);
K_OOPS(true);
}
}
/* Already copied *and* verified, no need to redo */
return z_impl_rtio_sqe_copy_in_get_handles(r, NULL, NULL, 0);
}
#include <zephyr/syscalls/rtio_sqe_copy_in_get_handles_mrsh.c>
static inline int z_vrfy_rtio_cqe_copy_out(struct rtio *r,
struct rtio_cqe *cqes,
size_t cqe_count,
k_timeout_t timeout)
{
K_OOPS(K_SYSCALL_OBJ(r, K_OBJ_RTIO));
K_OOPS(K_SYSCALL_MEMORY_ARRAY_WRITE(cqes, cqe_count, sizeof(struct rtio_cqe)));
return z_impl_rtio_cqe_copy_out(r, cqes, cqe_count, timeout);
}
#include <zephyr/syscalls/rtio_cqe_copy_out_mrsh.c>
static inline int z_vrfy_rtio_submit(struct rtio *r, uint32_t wait_count)
{
K_OOPS(K_SYSCALL_OBJ(r, K_OBJ_RTIO));
#ifdef CONFIG_RTIO_SUBMIT_SEM
K_OOPS(K_SYSCALL_OBJ(r->submit_sem, K_OBJ_SEM));
#endif
return z_impl_rtio_submit(r, wait_count);
}
#include <zephyr/syscalls/rtio_submit_mrsh.c>