blob: 00de793e6423a03c235a6678ee1d16d88edd8f4a [file] [log] [blame]
/*
* Copyright (c) 2014-2016, NVIDIA CORPORATION. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*/
#include <soc/tegra/ivc.h>
#define TEGRA_IVC_ALIGN 64
/*
* IVC channel reset protocol.
*
* Each end uses its tx_channel.state to indicate its synchronization state.
*/
enum tegra_ivc_state {
/*
* This value is zero for backwards compatibility with services that
* assume channels to be initially zeroed. Such channels are in an
* initially valid state, but cannot be asynchronously reset, and must
* maintain a valid state at all times.
*
* The transmitting end can enter the established state from the sync or
* ack state when it observes the receiving endpoint in the ack or
* established state, indicating that has cleared the counters in our
* rx_channel.
*/
TEGRA_IVC_STATE_ESTABLISHED = 0,
/*
* If an endpoint is observed in the sync state, the remote endpoint is
* allowed to clear the counters it owns asynchronously with respect to
* the current endpoint. Therefore, the current endpoint is no longer
* allowed to communicate.
*/
TEGRA_IVC_STATE_SYNC,
/*
* When the transmitting end observes the receiving end in the sync
* state, it can clear the w_count and r_count and transition to the ack
* state. If the remote endpoint observes us in the ack state, it can
* return to the established state once it has cleared its counters.
*/
TEGRA_IVC_STATE_ACK
};
/*
* This structure is divided into two-cache aligned parts, the first is only
* written through the tx.channel pointer, while the second is only written
* through the rx.channel pointer. This delineates ownership of the cache
* lines, which is critical to performance and necessary in non-cache coherent
* implementations.
*/
struct tegra_ivc_header {
union {
struct {
/* fields owned by the transmitting end */
u32 count;
u32 state;
};
u8 pad[TEGRA_IVC_ALIGN];
} tx;
union {
/* fields owned by the receiving end */
u32 count;
u8 pad[TEGRA_IVC_ALIGN];
} rx;
};
static inline void tegra_ivc_invalidate(struct tegra_ivc *ivc, dma_addr_t phys)
{
if (!ivc->peer)
return;
dma_sync_single_for_cpu(ivc->peer, phys, TEGRA_IVC_ALIGN,
DMA_FROM_DEVICE);
}
static inline void tegra_ivc_flush(struct tegra_ivc *ivc, dma_addr_t phys)
{
if (!ivc->peer)
return;
dma_sync_single_for_device(ivc->peer, phys, TEGRA_IVC_ALIGN,
DMA_TO_DEVICE);
}
static inline bool tegra_ivc_empty(struct tegra_ivc *ivc,
struct tegra_ivc_header *header)
{
/*
* This function performs multiple checks on the same values with
* security implications, so create snapshots with READ_ONCE() to
* ensure that these checks use the same values.
*/
u32 tx = READ_ONCE(header->tx.count);
u32 rx = READ_ONCE(header->rx.count);
/*
* Perform an over-full check to prevent denial of service attacks
* where a server could be easily fooled into believing that there's
* an extremely large number of frames ready, since receivers are not
* expected to check for full or over-full conditions.
*
* Although the channel isn't empty, this is an invalid case caused by
* a potentially malicious peer, so returning empty is safer, because
* it gives the impression that the channel has gone silent.
*/
if (tx - rx > ivc->num_frames)
return true;
return tx == rx;
}
static inline bool tegra_ivc_full(struct tegra_ivc *ivc,
struct tegra_ivc_header *header)
{
u32 tx = READ_ONCE(header->tx.count);
u32 rx = READ_ONCE(header->rx.count);
/*
* Invalid cases where the counters indicate that the queue is over
* capacity also appear full.
*/
return tx - rx >= ivc->num_frames;
}
static inline u32 tegra_ivc_available(struct tegra_ivc *ivc,
struct tegra_ivc_header *header)
{
u32 tx = READ_ONCE(header->tx.count);
u32 rx = READ_ONCE(header->rx.count);
/*
* This function isn't expected to be used in scenarios where an
* over-full situation can lead to denial of service attacks. See the
* comment in tegra_ivc_empty() for an explanation about special
* over-full considerations.
*/
return tx - rx;
}
static inline void tegra_ivc_advance_tx(struct tegra_ivc *ivc)
{
WRITE_ONCE(ivc->tx.channel->tx.count,
READ_ONCE(ivc->tx.channel->tx.count) + 1);
if (ivc->tx.position == ivc->num_frames - 1)
ivc->tx.position = 0;
else
ivc->tx.position++;
}
static inline void tegra_ivc_advance_rx(struct tegra_ivc *ivc)
{
WRITE_ONCE(ivc->rx.channel->rx.count,
READ_ONCE(ivc->rx.channel->rx.count) + 1);
if (ivc->rx.position == ivc->num_frames - 1)
ivc->rx.position = 0;
else
ivc->rx.position++;
}
static inline int tegra_ivc_check_read(struct tegra_ivc *ivc)
{
unsigned int offset = offsetof(struct tegra_ivc_header, tx.count);
/*
* tx.channel->state is set locally, so it is not synchronized with
* state from the remote peer. The remote peer cannot reset its
* transmit counters until we've acknowledged its synchronization
* request, so no additional synchronization is required because an
* asynchronous transition of rx.channel->state to
* TEGRA_IVC_STATE_ACK is not allowed.
*/
if (ivc->tx.channel->tx.state != TEGRA_IVC_STATE_ESTABLISHED)
return -ECONNRESET;
/*
* Avoid unnecessary invalidations when performing repeated accesses
* to an IVC channel by checking the old queue pointers first.
*
* Synchronization is only necessary when these pointers indicate
* empty or full.
*/
if (!tegra_ivc_empty(ivc, ivc->rx.channel))
return 0;
tegra_ivc_invalidate(ivc, ivc->rx.phys + offset);
if (tegra_ivc_empty(ivc, ivc->rx.channel))
return -ENOSPC;
return 0;
}
static inline int tegra_ivc_check_write(struct tegra_ivc *ivc)
{
unsigned int offset = offsetof(struct tegra_ivc_header, rx.count);
if (ivc->tx.channel->tx.state != TEGRA_IVC_STATE_ESTABLISHED)
return -ECONNRESET;
if (!tegra_ivc_full(ivc, ivc->tx.channel))
return 0;
tegra_ivc_invalidate(ivc, ivc->tx.phys + offset);
if (tegra_ivc_full(ivc, ivc->tx.channel))
return -ENOSPC;
return 0;
}
static void *tegra_ivc_frame_virt(struct tegra_ivc *ivc,
struct tegra_ivc_header *header,
unsigned int frame)
{
if (WARN_ON(frame >= ivc->num_frames))
return ERR_PTR(-EINVAL);
return (void *)(header + 1) + ivc->frame_size * frame;
}
static inline dma_addr_t tegra_ivc_frame_phys(struct tegra_ivc *ivc,
dma_addr_t phys,
unsigned int frame)
{
unsigned long offset;
offset = sizeof(struct tegra_ivc_header) + ivc->frame_size * frame;
return phys + offset;
}
static inline void tegra_ivc_invalidate_frame(struct tegra_ivc *ivc,
dma_addr_t phys,
unsigned int frame,
unsigned int offset,
size_t size)
{
if (!ivc->peer || WARN_ON(frame >= ivc->num_frames))
return;
phys = tegra_ivc_frame_phys(ivc, phys, frame) + offset;
dma_sync_single_for_cpu(ivc->peer, phys, size, DMA_FROM_DEVICE);
}
static inline void tegra_ivc_flush_frame(struct tegra_ivc *ivc,
dma_addr_t phys,
unsigned int frame,
unsigned int offset,
size_t size)
{
if (!ivc->peer || WARN_ON(frame >= ivc->num_frames))
return;
phys = tegra_ivc_frame_phys(ivc, phys, frame) + offset;
dma_sync_single_for_device(ivc->peer, phys, size, DMA_TO_DEVICE);
}
/* directly peek at the next frame rx'ed */
void *tegra_ivc_read_get_next_frame(struct tegra_ivc *ivc)
{
int err;
if (WARN_ON(ivc == NULL))
return ERR_PTR(-EINVAL);
err = tegra_ivc_check_read(ivc);
if (err < 0)
return ERR_PTR(err);
/*
* Order observation of ivc->rx.position potentially indicating new
* data before data read.
*/
smp_rmb();
tegra_ivc_invalidate_frame(ivc, ivc->rx.phys, ivc->rx.position, 0,
ivc->frame_size);
return tegra_ivc_frame_virt(ivc, ivc->rx.channel, ivc->rx.position);
}
EXPORT_SYMBOL(tegra_ivc_read_get_next_frame);
int tegra_ivc_read_advance(struct tegra_ivc *ivc)
{
unsigned int rx = offsetof(struct tegra_ivc_header, rx.count);
unsigned int tx = offsetof(struct tegra_ivc_header, tx.count);
int err;
/*
* No read barriers or synchronization here: the caller is expected to
* have already observed the channel non-empty. This check is just to
* catch programming errors.
*/
err = tegra_ivc_check_read(ivc);
if (err < 0)
return err;
tegra_ivc_advance_rx(ivc);
tegra_ivc_flush(ivc, ivc->rx.phys + rx);
/*
* Ensure our write to ivc->rx.position occurs before our read from
* ivc->tx.position.
*/
smp_mb();
/*
* Notify only upon transition from full to non-full. The available
* count can only asynchronously increase, so the worst possible
* side-effect will be a spurious notification.
*/
tegra_ivc_invalidate(ivc, ivc->rx.phys + tx);
if (tegra_ivc_available(ivc, ivc->rx.channel) == ivc->num_frames - 1)
ivc->notify(ivc, ivc->notify_data);
return 0;
}
EXPORT_SYMBOL(tegra_ivc_read_advance);
/* directly poke at the next frame to be tx'ed */
void *tegra_ivc_write_get_next_frame(struct tegra_ivc *ivc)
{
int err;
err = tegra_ivc_check_write(ivc);
if (err < 0)
return ERR_PTR(err);
return tegra_ivc_frame_virt(ivc, ivc->tx.channel, ivc->tx.position);
}
EXPORT_SYMBOL(tegra_ivc_write_get_next_frame);
/* advance the tx buffer */
int tegra_ivc_write_advance(struct tegra_ivc *ivc)
{
unsigned int tx = offsetof(struct tegra_ivc_header, tx.count);
unsigned int rx = offsetof(struct tegra_ivc_header, rx.count);
int err;
err = tegra_ivc_check_write(ivc);
if (err < 0)
return err;
tegra_ivc_flush_frame(ivc, ivc->tx.phys, ivc->tx.position, 0,
ivc->frame_size);
/*
* Order any possible stores to the frame before update of
* ivc->tx.position.
*/
smp_wmb();
tegra_ivc_advance_tx(ivc);
tegra_ivc_flush(ivc, ivc->tx.phys + tx);
/*
* Ensure our write to ivc->tx.position occurs before our read from
* ivc->rx.position.
*/
smp_mb();
/*
* Notify only upon transition from empty to non-empty. The available
* count can only asynchronously decrease, so the worst possible
* side-effect will be a spurious notification.
*/
tegra_ivc_invalidate(ivc, ivc->tx.phys + rx);
if (tegra_ivc_available(ivc, ivc->tx.channel) == 1)
ivc->notify(ivc, ivc->notify_data);
return 0;
}
EXPORT_SYMBOL(tegra_ivc_write_advance);
void tegra_ivc_reset(struct tegra_ivc *ivc)
{
unsigned int offset = offsetof(struct tegra_ivc_header, tx.count);
ivc->tx.channel->tx.state = TEGRA_IVC_STATE_SYNC;
tegra_ivc_flush(ivc, ivc->tx.phys + offset);
ivc->notify(ivc, ivc->notify_data);
}
EXPORT_SYMBOL(tegra_ivc_reset);
/*
* =======================================================
* IVC State Transition Table - see tegra_ivc_notified()
* =======================================================
*
* local remote action
* ----- ------ -----------------------------------
* SYNC EST <none>
* SYNC ACK reset counters; move to EST; notify
* SYNC SYNC reset counters; move to ACK; notify
* ACK EST move to EST; notify
* ACK ACK move to EST; notify
* ACK SYNC reset counters; move to ACK; notify
* EST EST <none>
* EST ACK <none>
* EST SYNC reset counters; move to ACK; notify
*
* ===============================================================
*/
int tegra_ivc_notified(struct tegra_ivc *ivc)
{
unsigned int offset = offsetof(struct tegra_ivc_header, tx.count);
enum tegra_ivc_state state;
/* Copy the receiver's state out of shared memory. */
tegra_ivc_invalidate(ivc, ivc->rx.phys + offset);
state = READ_ONCE(ivc->rx.channel->tx.state);
if (state == TEGRA_IVC_STATE_SYNC) {
offset = offsetof(struct tegra_ivc_header, tx.count);
/*
* Order observation of TEGRA_IVC_STATE_SYNC before stores
* clearing tx.channel.
*/
smp_rmb();
/*
* Reset tx.channel counters. The remote end is in the SYNC
* state and won't make progress until we change our state,
* so the counters are not in use at this time.
*/
ivc->tx.channel->tx.count = 0;
ivc->rx.channel->rx.count = 0;
ivc->tx.position = 0;
ivc->rx.position = 0;
/*
* Ensure that counters appear cleared before new state can be
* observed.
*/
smp_wmb();
/*
* Move to ACK state. We have just cleared our counters, so it
* is now safe for the remote end to start using these values.
*/
ivc->tx.channel->tx.state = TEGRA_IVC_STATE_ACK;
tegra_ivc_flush(ivc, ivc->tx.phys + offset);
/*
* Notify remote end to observe state transition.
*/
ivc->notify(ivc, ivc->notify_data);
} else if (ivc->tx.channel->tx.state == TEGRA_IVC_STATE_SYNC &&
state == TEGRA_IVC_STATE_ACK) {
offset = offsetof(struct tegra_ivc_header, tx.count);
/*
* Order observation of ivc_state_sync before stores clearing
* tx_channel.
*/
smp_rmb();
/*
* Reset tx.channel counters. The remote end is in the ACK
* state and won't make progress until we change our state,
* so the counters are not in use at this time.
*/
ivc->tx.channel->tx.count = 0;
ivc->rx.channel->rx.count = 0;
ivc->tx.position = 0;
ivc->rx.position = 0;
/*
* Ensure that counters appear cleared before new state can be
* observed.
*/
smp_wmb();
/*
* Move to ESTABLISHED state. We know that the remote end has
* already cleared its counters, so it is safe to start
* writing/reading on this channel.
*/
ivc->tx.channel->tx.state = TEGRA_IVC_STATE_ESTABLISHED;
tegra_ivc_flush(ivc, ivc->tx.phys + offset);
/*
* Notify remote end to observe state transition.
*/
ivc->notify(ivc, ivc->notify_data);
} else if (ivc->tx.channel->tx.state == TEGRA_IVC_STATE_ACK) {
offset = offsetof(struct tegra_ivc_header, tx.count);
/*
* At this point, we have observed the peer to be in either
* the ACK or ESTABLISHED state. Next, order observation of
* peer state before storing to tx.channel.
*/
smp_rmb();
/*
* Move to ESTABLISHED state. We know that we have previously
* cleared our counters, and we know that the remote end has
* cleared its counters, so it is safe to start writing/reading
* on this channel.
*/
ivc->tx.channel->tx.state = TEGRA_IVC_STATE_ESTABLISHED;
tegra_ivc_flush(ivc, ivc->tx.phys + offset);
/*
* Notify remote end to observe state transition.
*/
ivc->notify(ivc, ivc->notify_data);
} else {
/*
* There is no need to handle any further action. Either the
* channel is already fully established, or we are waiting for
* the remote end to catch up with our current state. Refer
* to the diagram in "IVC State Transition Table" above.
*/
}
if (ivc->tx.channel->tx.state != TEGRA_IVC_STATE_ESTABLISHED)
return -EAGAIN;
return 0;
}
EXPORT_SYMBOL(tegra_ivc_notified);
size_t tegra_ivc_align(size_t size)
{
return ALIGN(size, TEGRA_IVC_ALIGN);
}
EXPORT_SYMBOL(tegra_ivc_align);
unsigned tegra_ivc_total_queue_size(unsigned queue_size)
{
if (!IS_ALIGNED(queue_size, TEGRA_IVC_ALIGN)) {
pr_err("%s: queue_size (%u) must be %u-byte aligned\n",
__func__, queue_size, TEGRA_IVC_ALIGN);
return 0;
}
return queue_size + sizeof(struct tegra_ivc_header);
}
EXPORT_SYMBOL(tegra_ivc_total_queue_size);
static int tegra_ivc_check_params(unsigned long rx, unsigned long tx,
unsigned int num_frames, size_t frame_size)
{
BUILD_BUG_ON(!IS_ALIGNED(offsetof(struct tegra_ivc_header, tx.count),
TEGRA_IVC_ALIGN));
BUILD_BUG_ON(!IS_ALIGNED(offsetof(struct tegra_ivc_header, rx.count),
TEGRA_IVC_ALIGN));
BUILD_BUG_ON(!IS_ALIGNED(sizeof(struct tegra_ivc_header),
TEGRA_IVC_ALIGN));
if ((uint64_t)num_frames * (uint64_t)frame_size >= 0x100000000UL) {
pr_err("num_frames * frame_size overflows\n");
return -EINVAL;
}
if (!IS_ALIGNED(frame_size, TEGRA_IVC_ALIGN)) {
pr_err("frame size not adequately aligned: %zu\n", frame_size);
return -EINVAL;
}
/*
* The headers must at least be aligned enough for counters
* to be accessed atomically.
*/
if (!IS_ALIGNED(rx, TEGRA_IVC_ALIGN)) {
pr_err("IVC channel start not aligned: %#lx\n", rx);
return -EINVAL;
}
if (!IS_ALIGNED(tx, TEGRA_IVC_ALIGN)) {
pr_err("IVC channel start not aligned: %#lx\n", tx);
return -EINVAL;
}
if (rx < tx) {
if (rx + frame_size * num_frames > tx) {
pr_err("queue regions overlap: %#lx + %zx > %#lx\n",
rx, frame_size * num_frames, tx);
return -EINVAL;
}
} else {
if (tx + frame_size * num_frames > rx) {
pr_err("queue regions overlap: %#lx + %zx > %#lx\n",
tx, frame_size * num_frames, rx);
return -EINVAL;
}
}
return 0;
}
int tegra_ivc_init(struct tegra_ivc *ivc, struct device *peer, void *rx,
dma_addr_t rx_phys, void *tx, dma_addr_t tx_phys,
unsigned int num_frames, size_t frame_size,
void (*notify)(struct tegra_ivc *ivc, void *data),
void *data)
{
size_t queue_size;
int err;
if (WARN_ON(!ivc || !notify))
return -EINVAL;
/*
* All sizes that can be returned by communication functions should
* fit in an int.
*/
if (frame_size > INT_MAX)
return -E2BIG;
err = tegra_ivc_check_params((unsigned long)rx, (unsigned long)tx,
num_frames, frame_size);
if (err < 0)
return err;
queue_size = tegra_ivc_total_queue_size(num_frames * frame_size);
if (peer) {
ivc->rx.phys = dma_map_single(peer, rx, queue_size,
DMA_BIDIRECTIONAL);
if (dma_mapping_error(peer, ivc->rx.phys))
return -ENOMEM;
ivc->tx.phys = dma_map_single(peer, tx, queue_size,
DMA_BIDIRECTIONAL);
if (dma_mapping_error(peer, ivc->tx.phys)) {
dma_unmap_single(peer, ivc->rx.phys, queue_size,
DMA_BIDIRECTIONAL);
return -ENOMEM;
}
} else {
ivc->rx.phys = rx_phys;
ivc->tx.phys = tx_phys;
}
ivc->rx.channel = rx;
ivc->tx.channel = tx;
ivc->peer = peer;
ivc->notify = notify;
ivc->notify_data = data;
ivc->frame_size = frame_size;
ivc->num_frames = num_frames;
/*
* These values aren't necessarily correct until the channel has been
* reset.
*/
ivc->tx.position = 0;
ivc->rx.position = 0;
return 0;
}
EXPORT_SYMBOL(tegra_ivc_init);
void tegra_ivc_cleanup(struct tegra_ivc *ivc)
{
if (ivc->peer) {
size_t size = tegra_ivc_total_queue_size(ivc->num_frames *
ivc->frame_size);
dma_unmap_single(ivc->peer, ivc->rx.phys, size,
DMA_BIDIRECTIONAL);
dma_unmap_single(ivc->peer, ivc->tx.phys, size,
DMA_BIDIRECTIONAL);
}
}
EXPORT_SYMBOL(tegra_ivc_cleanup);