Hi all,
see attached a driver for the SIO coprocessor found on recent Apple
SoCs. This coprocessor provides general DMA services, it can feed
many peripherals but so far it seems it will only be useful for
audio output over HDMI/DisplayPort. So the driver here only supports
the DMA_CYCLIC mode of transactions with the focus being on audio.
There's a downstream prototype ALSA driver the DMA driver is being
tested against.
Some of the boilerplate code in implementing the dmaengine interface
was lifted from apple-admac.c. Among other things these two drivers
have in common that they implement the DMA_CYCLIC regime on top of
hardware/coprocessor layer supporting linear transactions only.
The binding schema saw two RFC rounds before and has a reviewed-by
from Rob.
https://lore.kernel.org/asahi/[email protected]
Best regards,
Martin
Martin Povišer (2):
dt-bindings: dma: apple,sio: Add schema
dmaengine: apple-sio: Add Apple SIO driver
.../devicetree/bindings/dma/apple,sio.yaml | 111 ++
MAINTAINERS | 2 +
drivers/dma/Kconfig | 10 +
drivers/dma/Makefile | 1 +
drivers/dma/apple-sio.c | 956 ++++++++++++++++++
5 files changed, 1080 insertions(+)
create mode 100644 Documentation/devicetree/bindings/dma/apple,sio.yaml
create mode 100644 drivers/dma/apple-sio.c
--
2.38.3
Add a dmaengine driver for the Apple SIO coprocessor found on Apple
SoCs where it provides DMA services. Have the driver support cyclic
transactions so that ALSA drivers can rely on it in audio output to
HDMI and DisplayPort.
Signed-off-by: Martin Povišer <[email protected]>
---
MAINTAINERS | 2 +
drivers/dma/Kconfig | 10 +
drivers/dma/Makefile | 1 +
drivers/dma/apple-sio.c | 956 ++++++++++++++++++++++++++++++++++++++++
4 files changed, 969 insertions(+)
create mode 100644 drivers/dma/apple-sio.c
diff --git a/MAINTAINERS b/MAINTAINERS
index 3be1bdfe8ecc..e65cf3d535ef 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1865,7 +1865,9 @@ M: Martin Povišer <[email protected]>
L: [email protected]
L: [email protected] (moderated for non-subscribers)
S: Maintained
+F: Documentation/devicetree/bindings/dma/apple,sio.yaml
F: Documentation/devicetree/bindings/sound/apple,*
+F: drivers/dma/apple-sio.c
F: sound/soc/apple/*
F: sound/soc/codecs/cs42l83-i2c.c
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 0f9832f0ef58..97b919835336 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -93,6 +93,16 @@ config APPLE_ADMAC
help
Enable support for Audio DMA Controller found on Apple Silicon SoCs.
+config APPLE_SIO
+ tristate "Apple SIO support"
+ depends on ARCH_APPLE || COMPILE_TEST
+ depends on APPLE_RTKIT
+ select DMA_ENGINE
+ default ARCH_APPLE
+ help
+ Enable support for the SIO coprocessor found on Apple Silicon SoCs
+ where it provides DMA services.
+
config AT_HDMAC
tristate "Atmel AHB DMA support"
depends on ARCH_AT91
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index a4fd1ce29510..adfecd893278 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -18,6 +18,7 @@ obj-$(CONFIG_AMBA_PL08X) += amba-pl08x.o
obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += ppc4xx/
obj-$(CONFIG_AMD_PTDMA) += ptdma/
obj-$(CONFIG_APPLE_ADMAC) += apple-admac.o
+obj-$(CONFIG_APPLE_SIO) += apple-sio.o
obj-$(CONFIG_AT_HDMAC) += at_hdmac.o
obj-$(CONFIG_AT_XDMAC) += at_xdmac.o
obj-$(CONFIG_AXI_DMAC) += dma-axi-dmac.o
diff --git a/drivers/dma/apple-sio.c b/drivers/dma/apple-sio.c
new file mode 100644
index 000000000000..87a27afaa22b
--- /dev/null
+++ b/drivers/dma/apple-sio.c
@@ -0,0 +1,956 @@
+// SPDX-License-Identifier: GPL-2.0-only OR MIT
+/*
+ * Driver for SIO coprocessor on t8103 (M1) and other Apple SoCs
+ *
+ * Copyright (C) The Asahi Linux Contributors
+ */
+
+#include <linux/bitfield.h>
+#include <linux/bits.h>
+#include <linux/completion.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/of_dma.h>
+#include <linux/soc/apple/rtkit.h>
+
+#include "dmaengine.h"
+
+#define NCHANNELS_MAX 0x80
+
+#define REG_CPU_CONTROL 0x44
+#define CPU_CONTROL_RUN BIT(4)
+
+#define SIOMSG_DATA GENMASK(63, 32)
+#define SIOMSG_TYPE GENMASK(23, 16)
+#define SIOMSG_PARAM GENMASK(31, 24)
+#define SIOMSG_TAG GENMASK(13, 8)
+#define SIOMSG_EP GENMASK(7, 0)
+
+#define EP_SIO 0x20
+
+#define MSG_START 0x2
+#define MSG_SETUP 0x3
+#define MSG_CONFIGURE 0x5
+#define MSG_ISSUE 0x6
+#define MSG_TERMINATE 0x8
+#define MSG_ACK 0x65
+#define MSG_NACK 0x66
+#define MSG_STARTED 0x67
+#define MSG_REPORT 0x68
+
+#define SIO_CALL_TIMEOUT_MS 100
+#define SIO_SHMEM_SIZE 0x1000
+#define SIO_NO_DESC_SLOTS 64
+
+/*
+ * There are two kinds of 'transaction descriptors' in play here.
+ *
+ * There's the struct sio_tx, and the struct dma_async_tx_descriptor embedded
+ * inside, which jointly represent a transaction to the dmaengine subsystem.
+ * At this time we only support those transactions to be cyclic.
+ *
+ * Then there are the coprocessor descriptors, which is what the coprocessor
+ * knows and understands. These don't seem to have a cyclic regime, so we can't
+ * map the dmaengine transaction on an exact coprocessor counterpart. Instead
+ * we continually queue up many coprocessor descriptors to implement a cyclic
+ * transaction.
+ *
+ * The number below is the maximum of how far ahead (how many) coprocessor
+ * descriptors we should be queuing up, per channel, for a cyclic transaction.
+ * Basically it's a made-up number.
+ */
+#define SIO_MAX_NINFLIGHT 4
+
+struct sio_coproc_desc {
+ u32 pad1;
+ u32 flag;
+ u64 unk;
+ u64 iova;
+ u64 size;
+ u64 pad2;
+ u64 pad3;
+} __packed;
+static_assert(sizeof(struct sio_coproc_desc) == 48);
+
+struct sio_shmem_chan_config {
+ u32 datashape;
+ u32 timeout;
+ u32 fifo;
+ u32 threshold;
+ u32 limit;
+} __packed;
+
+struct sio_data;
+struct sio_tx;
+
+struct sio_chan {
+ unsigned int no;
+ struct sio_data *host;
+ struct dma_chan chan;
+ struct tasklet_struct tasklet;
+ struct work_struct terminate_wq;
+
+ spinlock_t lock;
+ struct sio_tx *current_tx;
+ /*
+ * 'tx_cookie' is used for distinguishing between transactions from
+ * within tag ack/nack callbacks. Without it, we would have no way
+ * of knowing if the current transaction is the one the callback handler
+ * was installed for.
+ */
+ unsigned long tx_cookie;
+ int nperiod_acks;
+
+ /*
+ * We maintain a 'submitted' and 'issued' list mainly for interface
+ * correctness. Typical use of the driver (per channel) will be
+ * prepping, submitting and issuing a single cyclic transaction which
+ * will stay current until terminate_all is called.
+ */
+ struct list_head submitted;
+ struct list_head issued;
+
+ struct list_head to_free;
+};
+
+#define SIO_NTAGS 16
+
+typedef void (*sio_ack_callback)(struct sio_chan *, void *, bool);
+
+struct sio_data {
+ void __iomem *base;
+ struct dma_device dma;
+ struct device *dev;
+ struct apple_rtkit *rtk;
+ void *shmem;
+ struct sio_coproc_desc *shmem_desc_base;
+ unsigned long *desc_allocated;
+
+ struct sio_tagdata {
+ static_assert(sizeof(unsigned long) * 8 >= SIO_NTAGS);
+ unsigned long allocated;
+ int last_tag;
+
+ struct completion completions[SIO_NTAGS];
+ bool atomic[SIO_NTAGS];
+ bool acked[SIO_NTAGS];
+
+ sio_ack_callback ack_callback[SIO_NTAGS];
+ void *cookie[SIO_NTAGS];
+ } tags;
+
+ int nchannels;
+ struct sio_chan channels[];
+};
+
+struct sio_tx {
+ struct dma_async_tx_descriptor tx;
+ struct list_head node;
+ struct completion done;
+
+ bool terminated;
+ size_t period_len;
+ int nperiods;
+ int ninflight;
+ int next;
+
+ struct sio_coproc_desc *siodesc[];
+};
+
+static int sio_send_siomsg(struct sio_data *sio, u64 msg);
+static int sio_send_siomsg_atomic(struct sio_data *sio, u64 msg,
+ sio_ack_callback ack_callback,
+ void *cookie);
+static int sio_call(struct sio_data *sio, u64 msg);
+
+static struct sio_chan *to_sio_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sio_chan, chan);
+}
+
+static struct sio_tx *to_sio_tx(struct dma_async_tx_descriptor *tx)
+{
+ return container_of(tx, struct sio_tx, tx);
+}
+
+static int sio_alloc_tag(struct sio_data *sio)
+{
+ struct sio_tagdata *tags = &sio->tags;
+ int tag, i;
+
+ /*
+ * Because tag number 0 is special, the usable tag range
+ * is 1...(SIO_NTAGS - 1). So, to pick the next usable tag,
+ * we do modulo (SIO_NTAGS - 1) *then* plus one.
+ */
+
+#define SIO_USABLE_TAGS (SIO_NTAGS - 1)
+ tag = (READ_ONCE(tags->last_tag) % SIO_USABLE_TAGS) + 1;
+
+ for (i = 0; i < SIO_USABLE_TAGS; i++) {
+ if (!test_and_set_bit(tag, &tags->allocated))
+ break;
+
+ tag = (tag % SIO_USABLE_TAGS) + 1;
+ }
+
+ WRITE_ONCE(tags->last_tag, tag);
+
+ if (i < SIO_USABLE_TAGS)
+ return tag;
+ else
+ return -EBUSY;
+#undef SIO_USABLE_TAGS
+}
+
+static void sio_free_tag(struct sio_data *sio, int tag)
+{
+ struct sio_tagdata *tags = &sio->tags;
+
+ if (WARN_ON(tag >= SIO_NTAGS))
+ return;
+
+ tags->atomic[tag] = false;
+ tags->ack_callback[tag] = NULL;
+
+ WARN_ON(!test_and_clear_bit(tag, &tags->allocated));
+}
+
+static void sio_set_tag_atomic(struct sio_data *sio, int tag,
+ sio_ack_callback ack_callback,
+ void *cookie)
+{
+ struct sio_tagdata *tags = &sio->tags;
+
+ tags->atomic[tag] = true;
+ tags->ack_callback[tag] = ack_callback;
+ tags->cookie[tag] = cookie;
+}
+
+struct sio_coproc_desc *sio_alloc_desc(struct sio_data *sio)
+{
+ int i;
+
+ for (i = 0; i < SIO_NO_DESC_SLOTS; i++)
+ if (!test_and_set_bit(i, sio->desc_allocated))
+ return sio->shmem_desc_base + i;
+
+ return NULL;
+}
+
+void sio_free_desc(struct sio_data *sio, struct sio_coproc_desc *desc)
+{
+ clear_bit(desc - sio->shmem_desc_base, sio->desc_allocated);
+}
+
+int sio_coproc_desc_slot(struct sio_data *sio, struct sio_coproc_desc *desc)
+{
+ return (desc - sio->shmem_desc_base) * 4;
+}
+
+static enum dma_transfer_direction sio_chan_direction(int channo)
+{
+ /* Channel directions are fixed based on channel number */
+ return (channo & 1) ? DMA_DEV_TO_MEM : DMA_MEM_TO_DEV;
+}
+
+static dma_cookie_t sio_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sio_tx *siotx = to_sio_tx(tx);
+ struct sio_chan *siochan = to_sio_chan(tx->chan);
+ unsigned long flags;
+ dma_cookie_t cookie;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ cookie = dma_cookie_assign(tx);
+ list_add_tail(&siotx->node, &siochan->submitted);
+ spin_unlock_irqrestore(&siochan->lock, flags);
+
+ return cookie;
+}
+
+static int sio_tx_free(struct dma_async_tx_descriptor *tx)
+{
+ struct sio_data *sio = to_sio_chan(tx->chan)->host;
+ struct sio_tx *siotx = to_sio_tx(tx);
+ int i;
+
+ for (i = 0; i < siotx->nperiods; i++)
+ if (siotx->siodesc[i])
+ sio_free_desc(sio, siotx->siodesc[i]);
+ kfree(siotx);
+
+ return 0;
+}
+
+static struct dma_async_tx_descriptor *sio_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
+ size_t period_len, enum dma_transfer_direction direction,
+ unsigned long flags)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+ struct sio_tx *siotx = NULL;
+ int i, nperiods = buf_len / period_len;
+
+ if (direction != sio_chan_direction(siochan->no))
+ return NULL;
+
+ siotx = kzalloc(struct_size(siotx, siodesc, nperiods), GFP_NOWAIT);
+ if (!siotx)
+ return NULL;
+
+ init_completion(&siotx->done);
+ dma_async_tx_descriptor_init(&siotx->tx, chan);
+ siotx->period_len = period_len;
+ siotx->nperiods = nperiods;
+
+ for (i = 0; i < nperiods; i++) {
+ struct sio_coproc_desc *d;
+
+ siotx->siodesc[i] = d = sio_alloc_desc(siochan->host);
+ if (!d) {
+ sio_tx_free(&siotx->tx);
+ return NULL;
+ }
+
+ d->flag = 1; // not sure what's up with this
+ d->iova = buf_addr + period_len * i;
+ d->size = period_len;
+ }
+ dma_wmb();
+
+ siotx->tx.tx_submit = sio_tx_submit;
+ siotx->tx.desc_free = sio_tx_free;
+
+ return &siotx->tx;
+}
+
+static enum dma_status sio_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+ struct sio_tx *siotx;
+ enum dma_status ret;
+ unsigned long flags;
+ int periods_residue;
+ size_t residue;
+
+ ret = dma_cookie_status(chan, cookie, txstate);
+ if (ret == DMA_COMPLETE || !txstate)
+ return ret;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ siotx = siochan->current_tx;
+
+ if (siotx && siotx->tx.cookie == cookie) {
+ ret = DMA_IN_PROGRESS;
+ periods_residue = siotx->next - siotx->ninflight;
+ while (periods_residue < 0)
+ periods_residue += siotx->nperiods;
+ residue = (siotx->nperiods - periods_residue) * siotx->period_len;
+ } else {
+ ret = DMA_IN_PROGRESS;
+ residue = 0;
+ list_for_each_entry(siotx, &siochan->issued, node) {
+ if (siotx->tx.cookie == cookie) {
+ residue = siotx->period_len * siotx->nperiods;
+ break;
+ }
+ }
+ }
+ spin_unlock_irqrestore(&siochan->lock, flags);
+ dma_set_residue(txstate, residue);
+
+ return ret;
+}
+
+static bool sio_fill_in_locked(struct sio_chan *siochan);
+
+static void sio_handle_issue_ack(struct sio_chan *siochan, void *cookie, bool ok)
+{
+ unsigned long flags;
+ unsigned long tx_cookie = (unsigned long) cookie;
+ struct sio_tx *tx;
+
+ if (!ok) {
+ dev_err(siochan->host->dev, "nacked issue on chan %d\n", siochan->no);
+ return;
+ }
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ if (!siochan->current_tx || tx_cookie != siochan->tx_cookie ||
+ siochan->current_tx->terminated)
+ goto out;
+
+ tx = siochan->current_tx;
+ tx->next = (tx->next + 1) % tx->nperiods;
+ tx->ninflight++;
+ sio_fill_in_locked(siochan);
+
+out:
+ spin_unlock_irqrestore(&siochan->lock, flags);
+}
+
+static bool sio_fill_in_locked(struct sio_chan *siochan)
+{
+ struct sio_data *sio = siochan->host;
+ struct sio_tx *tx = siochan->current_tx;
+ struct sio_coproc_desc *d = tx->siodesc[tx->next];
+ int ret;
+
+ if (tx->ninflight >= SIO_MAX_NINFLIGHT || tx->terminated)
+ return false;
+
+ ret = sio_send_siomsg_atomic(sio, FIELD_PREP(SIOMSG_EP, siochan->no) |
+ FIELD_PREP(SIOMSG_TYPE, MSG_ISSUE) |
+ FIELD_PREP(SIOMSG_DATA, sio_coproc_desc_slot(sio, d)),
+ sio_handle_issue_ack, (void *) siochan->tx_cookie);
+ if (ret < 0)
+ dev_err_ratelimited(sio->dev, "can't issue on chan %d ninflight %d: %d\n",
+ siochan->no, tx->ninflight, ret);
+ return true;
+}
+
+static void sio_update_current_tx_locked(struct sio_chan *siochan)
+{
+ struct sio_tx *tx;
+
+ if (!list_empty(&siochan->issued) && !siochan->current_tx) {
+ tx = list_first_entry(&siochan->issued, struct sio_tx, node);
+ list_del(&tx->node);
+
+ siochan->current_tx = tx;
+ sio_fill_in_locked(siochan);
+ }
+}
+
+static void sio_issue_pending(struct dma_chan *chan)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+ unsigned long flags;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ list_splice_tail_init(&siochan->submitted, &siochan->issued);
+ sio_update_current_tx_locked(siochan);
+ spin_unlock_irqrestore(&siochan->lock, flags);
+}
+
+static int sio_terminate_all(struct dma_chan *chan)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+ unsigned long flags;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ if (siochan->current_tx) {
+ siochan->current_tx->terminated = true;
+ schedule_work(&siochan->terminate_wq);
+ }
+ list_splice_tail_init(&siochan->submitted, &siochan->to_free);
+ list_splice_tail_init(&siochan->issued, &siochan->to_free);
+ spin_unlock_irqrestore(&siochan->lock, flags);
+
+ return 0;
+}
+
+static void sio_terminate_work(struct work_struct *wq)
+{
+ struct sio_chan *siochan = container_of(wq, struct sio_chan, terminate_wq);
+ struct sio_tx *tx, *_tx;
+ unsigned long flags;
+ LIST_HEAD(to_free);
+ int ret;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ tx = siochan->current_tx;
+ list_splice_tail_init(&siochan->to_free, &to_free);
+ spin_unlock_irqrestore(&siochan->lock, flags);
+
+ if (WARN_ON(!tx))
+ return;
+
+ ret = sio_call(siochan->host, FIELD_PREP(SIOMSG_EP, siochan->no) |
+ FIELD_PREP(SIOMSG_TYPE, MSG_TERMINATE));
+ if (ret < 0)
+ dev_err(siochan->host->dev, "terminate call on chan %d failed: %d\n",
+ siochan->no, ret);
+
+ ret = wait_for_completion_timeout(&tx->done, msecs_to_jiffies(500));
+ if (!ret)
+ dev_err(siochan->host->dev, "terminate descriptor wait timed out\n");
+
+ tasklet_kill(&siochan->tasklet);
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ WARN_ON(siochan->current_tx != tx);
+ siochan->current_tx = NULL;
+ sio_update_current_tx_locked(siochan);
+ spin_unlock_irqrestore(&siochan->lock, flags);
+
+ sio_tx_free(&tx->tx);
+ list_for_each_entry_safe(tx, _tx, &to_free, node) {
+ list_del(&tx->node);
+ sio_tx_free(&tx->tx);
+ }
+}
+
+static void sio_synchronize(struct dma_chan *chan)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+
+ flush_work(&siochan->terminate_wq);
+}
+
+static int sio_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+
+ dma_cookie_init(&siochan->chan);
+ return 0;
+}
+
+static void sio_free_chan_resources(struct dma_chan *chan)
+{
+ sio_terminate_all(chan);
+ sio_synchronize(chan);
+}
+
+static struct dma_chan *sio_dma_of_xlate(struct of_phandle_args *dma_spec,
+ struct of_dma *ofdma)
+{
+ struct sio_data *sio = (struct sio_data *) ofdma->of_dma_data;
+ unsigned int index = dma_spec->args[0];
+
+ if (dma_spec->args_count != 1 || index >= sio->nchannels)
+ return ERR_PTR(-EINVAL);
+
+ return dma_get_slave_channel(&sio->channels[index].chan);
+}
+
+static void sio_rtk_crashed(void *cookie)
+{
+ struct sio_data *sio = cookie;
+
+ dev_err(sio->dev, "SIO down (crashed)");
+}
+
+static void sio_process_report(struct sio_chan *siochan)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ if (siochan->current_tx) {
+ struct sio_tx *tx = siochan->current_tx;
+
+ siochan->nperiod_acks++;
+ if (tx->ninflight)
+ tx->ninflight--;
+ if (!tx->terminated)
+ tasklet_schedule(&siochan->tasklet);
+ if (!sio_fill_in_locked(siochan) && !tx->ninflight)
+ complete(&tx->done);
+ }
+ spin_unlock_irqrestore(&siochan->lock, flags);
+}
+
+static void sio_recv_msg(void *cookie, u8 ep, u64 msg)
+{
+ struct sio_data *sio = cookie;
+ struct sio_tagdata *tags = &sio->tags;
+ u32 data;
+ u8 param, type, tag, sioep;
+
+ if (ep != EP_SIO)
+ goto unknown;
+
+ data = FIELD_GET(SIOMSG_DATA, msg);
+ param = FIELD_GET(SIOMSG_PARAM, msg);
+ type = FIELD_GET(SIOMSG_TYPE, msg);
+ tag = FIELD_GET(SIOMSG_TAG, msg);
+ sioep = FIELD_GET(SIOMSG_EP, msg);
+
+ switch (type) {
+ case MSG_STARTED:
+ dev_info(sio->dev, "SIO protocol v%u\n", data);
+ type = MSG_ACK; /* Pretend this is an ACK */
+ fallthrough;
+ case MSG_ACK:
+ case MSG_NACK:
+ if (WARN_ON(tag >= SIO_NTAGS))
+ break;
+
+ if (tags->atomic[tag]) {
+ sio_ack_callback callback = tags->ack_callback[tag];
+
+ if (callback && !WARN_ON(sioep >= sio->nchannels))
+ callback(&sio->channels[sioep],
+ tags->cookie[tag], type == MSG_ACK);
+ if (type == MSG_NACK)
+ dev_err(sio->dev, "got a NACK on channel %d\n", sioep);
+ sio_free_tag(sio, tag);
+ } else {
+ tags->acked[tag] = (type == MSG_ACK);
+ complete(&tags->completions[tag]);
+ }
+ break;
+
+ case MSG_REPORT:
+ if (WARN_ON(sioep >= sio->nchannels))
+ break;
+
+ sio_process_report(&sio->channels[sioep]);
+ break;
+
+ default:
+ goto unknown;
+ }
+ return;
+
+unknown:
+ dev_warn(sio->dev, "received unknown message: ep %x data %016llx\n",
+ ep, msg);
+}
+
+static int _sio_send_siomsg(struct sio_data *sio, u64 msg, bool atomic,
+ sio_ack_callback ack_callback, void *cookie)
+{
+ int tag, ret;
+
+ tag = sio_alloc_tag(sio);
+ if (tag < 0)
+ return tag;
+
+ if (atomic)
+ sio_set_tag_atomic(sio, tag, ack_callback, cookie);
+ else
+ reinit_completion(&sio->tags.completions[tag]);
+
+ msg &= ~SIOMSG_TAG;
+ msg |= FIELD_PREP(SIOMSG_TAG, tag);
+ ret = apple_rtkit_send_message(sio->rtk, EP_SIO, msg, NULL,
+ atomic);
+ if (ret < 0) {
+ sio_free_tag(sio, tag);
+ return ret;
+ }
+
+ return tag;
+}
+
+static int sio_send_siomsg(struct sio_data *sio, u64 msg)
+{
+ return _sio_send_siomsg(sio, msg, false, NULL, NULL);
+}
+
+static int sio_send_siomsg_atomic(struct sio_data *sio, u64 msg,
+ sio_ack_callback ack_callback,
+ void *cookie)
+{
+ return _sio_send_siomsg(sio, msg, true, ack_callback, cookie);
+}
+
+static int sio_call(struct sio_data *sio, u64 msg)
+{
+ int tag, ret;
+
+ tag = sio_send_siomsg(sio, msg);
+ if (tag < 0)
+ return tag;
+
+ ret = wait_for_completion_timeout(&sio->tags.completions[tag],
+ msecs_to_jiffies(SIO_CALL_TIMEOUT_MS));
+ if (!ret) {
+ dev_warn(sio->dev, "call %8x timed out\n", msg);
+ sio_free_tag(sio, tag);
+ return -ETIME;
+ }
+
+ ret = sio->tags.acked[tag];
+ sio_free_tag(sio, tag);
+
+ return ret;
+}
+
+static const struct apple_rtkit_ops sio_rtkit_ops = {
+ .crashed = sio_rtk_crashed,
+ .recv_message = sio_recv_msg,
+};
+
+static void sio_chan_tasklet(struct tasklet_struct *t)
+{
+ struct sio_chan *siochan = container_of(t, struct sio_chan, tasklet);
+ struct sio_tx *siotx;
+ struct dmaengine_desc_callback cb;
+ struct dmaengine_result tx_result;
+ unsigned long flags;
+ int nacks;
+
+ spin_lock_irqsave(&siochan->lock, flags);
+ siotx = siochan->current_tx;
+ nacks = siochan->nperiod_acks;
+ siochan->nperiod_acks = 0;
+ spin_unlock_irqrestore(&siochan->lock, flags);
+
+ if (!siotx || !nacks)
+ return;
+
+ tx_result.result = DMA_TRANS_NOERROR;
+ tx_result.residue = 0;
+
+ dmaengine_desc_get_callback(&siotx->tx, &cb);
+ while (nacks--)
+ dmaengine_desc_callback_invoke(&cb, &tx_result);
+}
+
+static int sio_device_config(struct dma_chan *chan,
+ struct dma_slave_config *config)
+{
+ struct sio_chan *siochan = to_sio_chan(chan);
+ struct sio_data *sio = siochan->host;
+ bool is_tx = sio_chan_direction(siochan->no) == DMA_MEM_TO_DEV;
+ struct sio_shmem_chan_config *cfg = sio->shmem;
+ int ret;
+
+ switch (is_tx ? config->dst_addr_width : config->src_addr_width) {
+ case DMA_SLAVE_BUSWIDTH_1_BYTE:
+ cfg->datashape = 0;
+ break;
+ case DMA_SLAVE_BUSWIDTH_2_BYTES:
+ cfg->datashape = 1;
+ break;
+ case DMA_SLAVE_BUSWIDTH_4_BYTES:
+ cfg->datashape = 2;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ cfg->fifo = 0x800;
+ cfg->limit = 0x800;
+ cfg->threshold = 0x800;
+ dma_wmb();
+
+ ret = sio_call(sio, FIELD_PREP(SIOMSG_TYPE, MSG_CONFIGURE) |
+ FIELD_PREP(SIOMSG_EP, siochan->no));
+
+ if (ret == 1)
+ ret = 0;
+ else if (ret == 0)
+ ret = -EINVAL;
+ return ret;
+}
+
+static int sio_alloc_shmem(struct sio_data *sio)
+{
+ dma_addr_t iova;
+ int err;
+
+ sio->shmem = dma_alloc_coherent(sio->dev, SIO_SHMEM_SIZE,
+ &iova, GFP_KERNEL);
+ if (!sio->shmem)
+ return -ENOMEM;
+
+ sio->shmem_desc_base = (struct sio_coproc_desc *) (sio->shmem + 56);
+ sio->desc_allocated = devm_kzalloc(sio->dev, SIO_NO_DESC_SLOTS / 32,
+ GFP_KERNEL);
+ if (!sio->desc_allocated)
+ return -ENOMEM;
+
+ err = sio_call(sio, FIELD_PREP(SIOMSG_TYPE, MSG_SETUP) |
+ FIELD_PREP(SIOMSG_PARAM, 1) |
+ FIELD_PREP(SIOMSG_DATA, iova >> 12));
+ if (err != 1) {
+ if (err == 0)
+ err = -EINVAL;
+ return err;
+ }
+
+ err = sio_call(sio, FIELD_PREP(SIOMSG_TYPE, MSG_SETUP) |
+ FIELD_PREP(SIOMSG_PARAM, 2) |
+ FIELD_PREP(SIOMSG_DATA, SIO_SHMEM_SIZE));
+ if (err != 1) {
+ if (err == 0)
+ err = -EINVAL;
+ return err;
+ }
+
+ return 0;
+}
+
+static int sio_send_dt_params(struct sio_data *sio)
+{
+ struct device_node *np = sio->dev->of_node;
+ const char *propname = "apple,sio-firmware-params";
+ int nparams, err, i;
+
+ nparams = of_property_count_u32_elems(np, propname);
+ if (nparams < 0) {
+ err = nparams;
+ goto badprop;
+ }
+
+ for (i = 0; i < nparams / 2; i++) {
+ u32 key, val;
+
+ err = of_property_read_u32_index(np, propname, 2 * i, &key);
+ if (err)
+ goto badprop;
+ err = of_property_read_u32_index(np, propname, 2 * i + 1, &val);
+ if (err)
+ goto badprop;
+
+ err = sio_call(sio, FIELD_PREP(SIOMSG_TYPE, MSG_SETUP) |
+ FIELD_PREP(SIOMSG_PARAM, key & 0xff) |
+ FIELD_PREP(SIOMSG_EP, key >> 8) |
+ FIELD_PREP(SIOMSG_DATA, val));
+ if (err < 1) {
+ if (err == 0)
+ err = -ENXIO;
+ return dev_err_probe(sio->dev, err, "sending SIO parameter %#x value %#x\n",
+ key, val);
+ }
+ }
+
+ return 0;
+
+badprop:
+ return dev_err_probe(sio->dev, err, "failed to read '%s'\n", propname);
+}
+
+static int sio_probe(struct platform_device *pdev)
+{
+ struct device_node *np = pdev->dev.of_node;
+ struct sio_data *sio;
+ struct dma_device *dma;
+ int nchannels;
+ int err, i;
+
+ err = of_property_read_u32(np, "dma-channels", &nchannels);
+ if (err || nchannels > NCHANNELS_MAX)
+ return dev_err_probe(&pdev->dev, -EINVAL,
+ "missing or invalid dma-channels property\n");
+
+ sio = devm_kzalloc(&pdev->dev, struct_size(sio, channels, nchannels), GFP_KERNEL);
+ if (!sio)
+ return -ENOMEM;
+
+ platform_set_drvdata(pdev, sio);
+ sio->dev = &pdev->dev;
+ sio->nchannels = nchannels;
+
+ sio->base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(sio->base))
+ return PTR_ERR(sio->base);
+
+ sio->rtk = devm_apple_rtkit_init(&pdev->dev, sio, NULL, 0, &sio_rtkit_ops);
+ if (IS_ERR(sio->rtk))
+ return dev_err_probe(&pdev->dev, PTR_ERR(sio->rtk),
+ "couldn't initialize rtkit\n");
+ for (i = 1; i < SIO_NTAGS; i++)
+ init_completion(&sio->tags.completions[i]);
+
+ dma = &sio->dma;
+ dma_cap_set(DMA_PRIVATE, dma->cap_mask);
+ dma_cap_set(DMA_CYCLIC, dma->cap_mask);
+
+ dma->dev = &pdev->dev;
+ dma->device_alloc_chan_resources = sio_alloc_chan_resources;
+ dma->device_free_chan_resources = sio_free_chan_resources;
+ dma->device_tx_status = sio_tx_status;
+ dma->device_issue_pending = sio_issue_pending;
+ dma->device_terminate_all = sio_terminate_all;
+ dma->device_synchronize = sio_synchronize;
+ dma->device_prep_dma_cyclic = sio_prep_dma_cyclic;
+ dma->device_config = sio_device_config;
+
+ dma->directions = BIT(DMA_MEM_TO_DEV);
+ dma->residue_granularity = DMA_RESIDUE_GRANULARITY_SEGMENT;
+ dma->dst_addr_widths = BIT(DMA_SLAVE_BUSWIDTH_1_BYTE) |
+ BIT(DMA_SLAVE_BUSWIDTH_2_BYTES) |
+ BIT(DMA_SLAVE_BUSWIDTH_4_BYTES);
+
+ INIT_LIST_HEAD(&dma->channels);
+ for (i = 0; i < nchannels; i++) {
+ struct sio_chan *siochan = &sio->channels[i];
+
+ siochan->host = sio;
+ siochan->no = i;
+ siochan->chan.device = &sio->dma;
+ spin_lock_init(&siochan->lock);
+ INIT_LIST_HEAD(&siochan->submitted);
+ INIT_LIST_HEAD(&siochan->issued);
+ INIT_LIST_HEAD(&siochan->to_free);
+ list_add_tail(&siochan->chan.device_node, &dma->channels);
+ tasklet_setup(&siochan->tasklet, sio_chan_tasklet);
+ INIT_WORK(&siochan->terminate_wq, sio_terminate_work);
+ }
+
+ writel(CPU_CONTROL_RUN, sio->base + REG_CPU_CONTROL);
+
+ err = apple_rtkit_boot(sio->rtk);
+ if (err)
+ return dev_err_probe(&pdev->dev, err, "SIO did not boot\n");
+
+ err = apple_rtkit_start_ep(sio->rtk, EP_SIO);
+ if (err)
+ return dev_err_probe(&pdev->dev, err, "starting SIO endpoint\n");
+
+ err = sio_call(sio, FIELD_PREP(SIOMSG_TYPE, MSG_START));
+ if (err < 1) {
+ if (err == 0)
+ err = -ENXIO;
+ return dev_err_probe(&pdev->dev, err, "starting SIO service\n");
+ }
+
+ err = sio_send_dt_params(sio);
+ if (err < 0)
+ return dev_err_probe(&pdev->dev, err, "failed to send boot-up parameters\n");
+
+ err = sio_alloc_shmem(sio);
+ if (err < 0)
+ return err;
+
+ err = dma_async_device_register(&sio->dma);
+ if (err)
+ return dev_err_probe(&pdev->dev, err, "failed to register DMA device\n");
+
+ err = of_dma_controller_register(pdev->dev.of_node, sio_dma_of_xlate, sio);
+ if (err) {
+ dma_async_device_unregister(&sio->dma);
+ return dev_err_probe(&pdev->dev, err, "failed to register with OF\n");
+ }
+
+ return 0;
+}
+
+static int sio_remove(struct platform_device *pdev)
+{
+ struct sio_data *sio = platform_get_drvdata(pdev);
+
+ of_dma_controller_free(pdev->dev.of_node);
+ dma_async_device_unregister(&sio->dma);
+ return 0;
+}
+
+static const struct of_device_id sio_of_match[] = {
+ { .compatible = "apple,sio", },
+ { }
+};
+MODULE_DEVICE_TABLE(of, sio_of_match);
+
+static struct platform_driver apple_sio_driver = {
+ .driver = {
+ .name = "apple-sio",
+ .of_match_table = sio_of_match,
+ },
+ .probe = sio_probe,
+ .remove = sio_remove,
+};
+module_platform_driver(apple_sio_driver);
+
+MODULE_AUTHOR("Martin Povišer <[email protected]>");
+MODULE_DESCRIPTION("Driver for SIO coprocessor on Apple SoCs");
+MODULE_LICENSE("Dual MIT/GPL");
--
2.38.3
Describe the SIO coprocessor which serves as pretend DMA controller on
recent Apple platforms.
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Martin Povišer <[email protected]>
---
.../devicetree/bindings/dma/apple,sio.yaml | 111 ++++++++++++++++++
1 file changed, 111 insertions(+)
create mode 100644 Documentation/devicetree/bindings/dma/apple,sio.yaml
diff --git a/Documentation/devicetree/bindings/dma/apple,sio.yaml b/Documentation/devicetree/bindings/dma/apple,sio.yaml
new file mode 100644
index 000000000000..0e3780ad9dd7
--- /dev/null
+++ b/Documentation/devicetree/bindings/dma/apple,sio.yaml
@@ -0,0 +1,111 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/dma/apple,sio.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Apple SIO Coprocessor
+
+description:
+ SIO is a coprocessor on Apple M1 and later chips (and maybe also on earlier
+ chips). Its role is to offload SPI, UART and DisplayPort audio transfers,
+ being a pretend DMA controller.
+
+maintainers:
+ - Martin Povišer <[email protected]>
+
+allOf:
+ - $ref: dma-controller.yaml#
+
+properties:
+ compatible:
+ items:
+ - enum:
+ - apple,t6000-sio
+ - apple,t8103-sio
+ - const: apple,sio
+
+ reg:
+ maxItems: 1
+
+ '#dma-cells':
+ const: 1
+ description:
+ DMA clients specify a single cell that corresponds to the RTKit endpoint
+ number used for arranging the transfers in question
+
+ dma-channels:
+ maximum: 128
+
+ mboxes:
+ maxItems: 1
+
+ iommus:
+ maxItems: 1
+
+ power-domains:
+ maxItems: 1
+
+ memory-region:
+ minItems: 2
+ maxItems: 8
+ description:
+ A number of references to reserved memory regions among which are the DATA/TEXT
+ sections of coprocessor executable firmware and also auxiliary firmware data
+ describing the available DMA-enabled peripherals
+
+ apple,sio-firmware-params:
+ $ref: /schemas/types.yaml#/definitions/uint32-array
+ description: |
+ Parameters in the form of opaque key/value pairs that are to be sent to the SIO
+ coprocesssor once it boots. These parameters can point into the reserved memory
+ regions (in device address space).
+
+ Note that unlike Apple's firmware, we treat the parameters, and the data they
+ refer to, as opaque. Apple embed short data blobs into their SIO devicetree node
+ that describe the DMA-enabled peripherals (presumably with defined semantics).
+ Their driver processes those blobs and sets up data structure in mapped device
+ memory, then references this memory in the parameters sent to the SIO. At the
+ level of description we are opting for in this binding, we assume the job of
+ constructing those data structures has been done in advance, leaving behind an
+ opaque list of key/value parameter pairs to be sent by a prospective driver.
+
+ This approach is chosen for two reasons:
+
+ - It means we don't need to try to understand the semantics of Apple's blobs
+ as long as we know the transformation we need to do from Apple's devicetree
+ data to SIO data (which can be shoved away into a loader). It also means the
+ semantics of Apple's blobs (or of something to replace them) need not be part
+ of the binding and be kept up with Apple's firmware changes in the future.
+
+ - It leaves less work for the driver attaching on this binding. Instead the work
+ is done upfront in the loader which can be better suited for keeping up with
+ Apple's firmware changes.
+
+required:
+ - compatible
+ - reg
+ - '#dma-cells'
+ - dma-channels
+ - mboxes
+ - iommus
+ - power-domains
+
+additionalProperties: false
+
+examples:
+ - |
+ sio: dma-controller@36400000 {
+ compatible = "apple,t8103-sio", "apple,sio";
+ reg = <0x36400000 0x8000>;
+ dma-channels = <128>;
+ #dma-cells = <1>;
+ mboxes = <&sio_mbox>;
+ iommus = <&sio_dart 0>;
+ power-domains = <&ps_sio_cpu>;
+ memory-region = <&sio_text>, <&sio_data>,
+ <&sio_auxdata1>, <&sio_auxdata2>; /* Filled by loader */
+ apple,sio-firmware-params = <0xb 0x10>, <0xc 0x1b80>, <0xf 0x14>,
+ <0x10 0x1e000>, <0x30d 0x34>, <0x30e 0x4000>,
+ <0x1a 0x38>, <0x1b 0x50>; /* Filled by loader */
+ };
--
2.38.3
On 12/07/2023 15:38, Martin Povišer wrote:
> Hi all,
>
> see attached a driver for the SIO coprocessor found on recent Apple
> SoCs. This coprocessor provides general DMA services, it can feed
> many peripherals but so far it seems it will only be useful for
> audio output over HDMI/DisplayPort. So the driver here only supports
> the DMA_CYCLIC mode of transactions with the focus being on audio.
> There's a downstream prototype ALSA driver the DMA driver is being
> tested against.
>
> Some of the boilerplate code in implementing the dmaengine interface
> was lifted from apple-admac.c. Among other things these two drivers
> have in common that they implement the DMA_CYCLIC regime on top of
> hardware/coprocessor layer supporting linear transactions only.
>
> The binding schema saw two RFC rounds before and has a reviewed-by
> from Rob.
> https://lore.kernel.org/asahi/[email protected]
Thank you for explanation. Then this is v3, not v1.
No need for resending, but if it happens, consider naming it v4. :)
Best regards,
Krzysztof
On 12-07-23, 15:38, Martin Povišer wrote:
> +struct sio_chan {
> + unsigned int no;
> + struct sio_data *host;
> + struct dma_chan chan;
> + struct tasklet_struct tasklet;
> + struct work_struct terminate_wq;
> +
> + spinlock_t lock;
> + struct sio_tx *current_tx;
> + /*
> + * 'tx_cookie' is used for distinguishing between transactions from
> + * within tag ack/nack callbacks. Without it, we would have no way
> + * of knowing if the current transaction is the one the callback handler
> + * was installed for.
not sure what you mean by here.. I dont see why you would need to store
cookie here, care to explain?
> + */
> + unsigned long tx_cookie;
> + int nperiod_acks;
> +
> + /*
> + * We maintain a 'submitted' and 'issued' list mainly for interface
> + * correctness. Typical use of the driver (per channel) will be
> + * prepping, submitting and issuing a single cyclic transaction which
> + * will stay current until terminate_all is called.
> + */
> + struct list_head submitted;
> + struct list_head issued;
> +
> + struct list_head to_free;
can you use virt_dma_chan, that should simplify list handling etc
> +};
> +
> +#define SIO_NTAGS 16
> +
> +typedef void (*sio_ack_callback)(struct sio_chan *, void *, bool);
any reason not to use dmaengine callbacks?
> +static int sio_alloc_tag(struct sio_data *sio)
> +{
> + struct sio_tagdata *tags = &sio->tags;
> + int tag, i;
> +
> + /*
> + * Because tag number 0 is special, the usable tag range
> + * is 1...(SIO_NTAGS - 1). So, to pick the next usable tag,
> + * we do modulo (SIO_NTAGS - 1) *then* plus one.
> + */
> +
> +#define SIO_USABLE_TAGS (SIO_NTAGS - 1)
> + tag = (READ_ONCE(tags->last_tag) % SIO_USABLE_TAGS) + 1;
> +
> + for (i = 0; i < SIO_USABLE_TAGS; i++) {
> + if (!test_and_set_bit(tag, &tags->allocated))
> + break;
> +
> + tag = (tag % SIO_USABLE_TAGS) + 1;
> + }
> +
> + WRITE_ONCE(tags->last_tag, tag);
> +
> + if (i < SIO_USABLE_TAGS)
> + return tag;
> + else
> + return -EBUSY;
> +#undef SIO_USABLE_TAGS
> +}
can you use kernel mechanisms like ida to alloc and free the tags...
> +static struct dma_async_tx_descriptor *sio_prep_dma_cyclic(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_transfer_direction direction,
> + unsigned long flags)
> +{
> + struct sio_chan *siochan = to_sio_chan(chan);
> + struct sio_tx *siotx = NULL;
> + int i, nperiods = buf_len / period_len;
> +
> + if (direction != sio_chan_direction(siochan->no))
> + return NULL;
> +
> + siotx = kzalloc(struct_size(siotx, siodesc, nperiods), GFP_NOWAIT);
> + if (!siotx)
> + return NULL;
> +
> + init_completion(&siotx->done);
> + dma_async_tx_descriptor_init(&siotx->tx, chan);
> + siotx->period_len = period_len;
> + siotx->nperiods = nperiods;
> +
> + for (i = 0; i < nperiods; i++) {
> + struct sio_coproc_desc *d;
> +
> + siotx->siodesc[i] = d = sio_alloc_desc(siochan->host);
> + if (!d) {
> + sio_tx_free(&siotx->tx);
> + return NULL;
> + }
> +
> + d->flag = 1; // not sure what's up with this
> + d->iova = buf_addr + period_len * i;
> + d->size = period_len;
> + }
> + dma_wmb();
why use barrier here? and to what purpose..
--
~Vinod
Hi Vinod!
> On 1. 8. 2023, at 20:14, Vinod Koul <[email protected]> wrote:
>
> On 12-07-23, 15:38, Martin Povišer wrote:
>
>> +struct sio_chan {
>> + unsigned int no;
>> + struct sio_data *host;
>> + struct dma_chan chan;
>> + struct tasklet_struct tasklet;
>> + struct work_struct terminate_wq;
>> +
>> + spinlock_t lock;
>> + struct sio_tx *current_tx;
>> + /*
>> + * 'tx_cookie' is used for distinguishing between transactions from
>> + * within tag ack/nack callbacks. Without it, we would have no way
>> + * of knowing if the current transaction is the one the callback handler
>> + * was installed for.
>
> not sure what you mean by here.. I dont see why you would need to store
> cookie here, care to explain?
I could have clarified this is not meant to be the dmaengine cookie, just
a driver-level cookie to address a race between
a dmaengine user calling terminate_all to terminate a running
cyclic transaction, then issuing a new one
on one hand, and
the coprocessor acking the issuing of one of the coprocessor
transactions that correspond to the first dmaengine transaction
on the other hand. With the cookie the driver should not get confused
about which dmaengine transaction the ACK belongs to, since if `current_tx`
changed in the meantime the cookie won’t match.
But now that I look at it... huh, I never increment that `tx_cookie` field!
I don’t know if I have considered using the dmaengine cookie to the same
effect. Maybe we can do that, I see how that would be much desirable.
>> + */
>> + unsigned long tx_cookie;
>> + int nperiod_acks;
>> +
>> + /*
>> + * We maintain a 'submitted' and 'issued' list mainly for interface
>> + * correctness. Typical use of the driver (per channel) will be
>> + * prepping, submitting and issuing a single cyclic transaction which
>> + * will stay current until terminate_all is called.
>> + */
>> + struct list_head submitted;
>> + struct list_head issued;
>> +
>> + struct list_head to_free;
>
> can you use virt_dma_chan, that should simplify list handling etc
I looked into that when I wrote the sister driver apple-admac.c, I don’t
remember anymore why I decided against it, and I don’t think it came up
during review. Now that this driver is done, I hope we can take it as is.
There’s some benefit from the drivers having a similar structure, I sent
one or two fixes to apple-admac for things I found out because I was
writing this other driver.
>> +};
>> +
>> +#define SIO_NTAGS 16
>> +
>> +typedef void (*sio_ack_callback)(struct sio_chan *, void *, bool);
>
> any reason not to use dmaengine callbacks?
Not sure what dmaengine callback you mean here. This callback means
the coprocessor acked a tag, not sure how we can fit something dmaengine
onto it.
>> +static int sio_alloc_tag(struct sio_data *sio)
>> +{
>> + struct sio_tagdata *tags = &sio->tags;
>> + int tag, i;
>> +
>> + /*
>> + * Because tag number 0 is special, the usable tag range
>> + * is 1...(SIO_NTAGS - 1). So, to pick the next usable tag,
>> + * we do modulo (SIO_NTAGS - 1) *then* plus one.
>> + */
>> +
>> +#define SIO_USABLE_TAGS (SIO_NTAGS - 1)
>> + tag = (READ_ONCE(tags->last_tag) % SIO_USABLE_TAGS) + 1;
>> +
>> + for (i = 0; i < SIO_USABLE_TAGS; i++) {
>> + if (!test_and_set_bit(tag, &tags->allocated))
>> + break;
>> +
>> + tag = (tag % SIO_USABLE_TAGS) + 1;
>> + }
>> +
>> + WRITE_ONCE(tags->last_tag, tag);
>> +
>> + if (i < SIO_USABLE_TAGS)
>> + return tag;
>> + else
>> + return -EBUSY;
>> +#undef SIO_USABLE_TAGS
>> +}
>
> can you use kernel mechanisms like ida to alloc and free the tags...
I can look into that.
>> +static struct dma_async_tx_descriptor *sio_prep_dma_cyclic(
>> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
>> + size_t period_len, enum dma_transfer_direction direction,
>> + unsigned long flags)
>> +{
>> + struct sio_chan *siochan = to_sio_chan(chan);
>> + struct sio_tx *siotx = NULL;
>> + int i, nperiods = buf_len / period_len;
>> +
>> + if (direction != sio_chan_direction(siochan->no))
>> + return NULL;
>> +
>> + siotx = kzalloc(struct_size(siotx, siodesc, nperiods), GFP_NOWAIT);
>> + if (!siotx)
>> + return NULL;
>> +
>> + init_completion(&siotx->done);
>> + dma_async_tx_descriptor_init(&siotx->tx, chan);
>> + siotx->period_len = period_len;
>> + siotx->nperiods = nperiods;
>> +
>> + for (i = 0; i < nperiods; i++) {
>> + struct sio_coproc_desc *d;
>> +
>> + siotx->siodesc[i] = d = sio_alloc_desc(siochan->host);
>> + if (!d) {
>> + sio_tx_free(&siotx->tx);
>> + return NULL;
>> + }
>> +
>> + d->flag = 1; // not sure what's up with this
>> + d->iova = buf_addr + period_len * i;
>> + d->size = period_len;
>> + }
>> + dma_wmb();
>
> why use barrier here? and to what purpose..
Few lines above we are modifying a shared memory buffer that’s mapped into
the coprocessor’s address space (it’s what `d` points to).
> --
> ~Vinod
>
Best regards, Martin
> On 1. 8. 2023, at 23:55, Martin Povišer <[email protected]> wrote:
>
> Hi Vinod!
>
>> On 1. 8. 2023, at 20:14, Vinod Koul <[email protected]> wrote:
>>
>> On 12-07-23, 15:38, Martin Povišer wrote:
>>
>>> +struct sio_chan {
>>> + unsigned int no;
>>> + struct sio_data *host;
>>> + struct dma_chan chan;
>>> + struct tasklet_struct tasklet;
>>> + struct work_struct terminate_wq;
>>> +
>>> + spinlock_t lock;
>>> + struct sio_tx *current_tx;
>>> + /*
>>> + * 'tx_cookie' is used for distinguishing between transactions from
>>> + * within tag ack/nack callbacks. Without it, we would have no way
>>> + * of knowing if the current transaction is the one the callback handler
>>> + * was installed for.
>>
>> not sure what you mean by here.. I dont see why you would need to store
>> cookie here, care to explain?
>
> I could have clarified this is not meant to be the dmaengine cookie, just
> a driver-level cookie to address a race between
>
> a dmaengine user calling terminate_all to terminate a running
> cyclic transaction, then issuing a new one
>
> on one hand, and
>
> the coprocessor acking the issuing of one of the coprocessor
> transactions that correspond to the first dmaengine transaction
>
> on the other hand. With the cookie the driver should not get confused
> about which dmaengine transaction the ACK belongs to, since if `current_tx`
> changed in the meantime the cookie won’t match.
>
> But now that I look at it... huh, I never increment that `tx_cookie` field!
> I don’t know if I have considered using the dmaengine cookie to the same
> effect. Maybe we can do that, I see how that would be much desirable.
Indeed nothing is stopping us from matching on the dmaengine cookie to
address the race, so I will be dropping this `tx_cookie` field in v2.
>>> +static int sio_alloc_tag(struct sio_data *sio)
>>> +{
>>> + struct sio_tagdata *tags = &sio->tags;
>>> + int tag, i;
>>> +
>>> + /*
>>> + * Because tag number 0 is special, the usable tag range
>>> + * is 1...(SIO_NTAGS - 1). So, to pick the next usable tag,
>>> + * we do modulo (SIO_NTAGS - 1) *then* plus one.
>>> + */
>>> +
>>> +#define SIO_USABLE_TAGS (SIO_NTAGS - 1)
>>> + tag = (READ_ONCE(tags->last_tag) % SIO_USABLE_TAGS) + 1;
>>> +
>>> + for (i = 0; i < SIO_USABLE_TAGS; i++) {
>>> + if (!test_and_set_bit(tag, &tags->allocated))
>>> + break;
>>> +
>>> + tag = (tag % SIO_USABLE_TAGS) + 1;
>>> + }
>>> +
>>> + WRITE_ONCE(tags->last_tag, tag);
>>> +
>>> + if (i < SIO_USABLE_TAGS)
>>> + return tag;
>>> + else
>>> + return -EBUSY;
>>> +#undef SIO_USABLE_TAGS
>>> +}
>>
>> can you use kernel mechanisms like ida to alloc and free the tags...
>
> I can look into that.
Documentation says IDA is deprecated in favour of Xarray, both look
like they serve to associate a pointer with an ID. I think neither
structure beats a simple bitfield and a static array for the per-tag
data. Agree?
Martin
On 01-08-23, 23:55, Martin Povišer wrote:
> > can you use virt_dma_chan, that should simplify list handling etc
>
> I looked into that when I wrote the sister driver apple-admac.c, I don’t
> remember anymore why I decided against it, and I don’t think it came up
> during review. Now that this driver is done, I hope we can take it as is.
>
> There’s some benefit from the drivers having a similar structure, I sent
> one or two fixes to apple-admac for things I found out because I was
> writing this other driver.
And this would be a chance to covert the other one and get rid of list
handling code in that driver as well
>
> >> +};
> >> +
> >> +#define SIO_NTAGS 16
> >> +
> >> +typedef void (*sio_ack_callback)(struct sio_chan *, void *, bool);
> >
> > any reason not to use dmaengine callbacks?
>
> Not sure what dmaengine callback you mean here. This callback means
> the coprocessor acked a tag, not sure how we can fit something dmaengine
> onto it.
Okay lets understand, how is this one used
--
~Vinod
On 03-08-23, 10:32, Martin Povišer wrote:
> >>> +static int sio_alloc_tag(struct sio_data *sio)
> >>> +{
> >>> + struct sio_tagdata *tags = &sio->tags;
> >>> + int tag, i;
> >>> +
> >>> + /*
> >>> + * Because tag number 0 is special, the usable tag range
> >>> + * is 1...(SIO_NTAGS - 1). So, to pick the next usable tag,
> >>> + * we do modulo (SIO_NTAGS - 1) *then* plus one.
> >>> + */
> >>> +
> >>> +#define SIO_USABLE_TAGS (SIO_NTAGS - 1)
> >>> + tag = (READ_ONCE(tags->last_tag) % SIO_USABLE_TAGS) + 1;
> >>> +
> >>> + for (i = 0; i < SIO_USABLE_TAGS; i++) {
> >>> + if (!test_and_set_bit(tag, &tags->allocated))
> >>> + break;
> >>> +
> >>> + tag = (tag % SIO_USABLE_TAGS) + 1;
> >>> + }
> >>> +
> >>> + WRITE_ONCE(tags->last_tag, tag);
> >>> +
> >>> + if (i < SIO_USABLE_TAGS)
> >>> + return tag;
> >>> + else
> >>> + return -EBUSY;
> >>> +#undef SIO_USABLE_TAGS
> >>> +}
> >>
> >> can you use kernel mechanisms like ida to alloc and free the tags...
> >
> > I can look into that.
>
> Documentation says IDA is deprecated in favour of Xarray, both look
> like they serve to associate a pointer with an ID. I think neither
> structure beats a simple bitfield and a static array for the per-tag
> data. Agree?
yeah xarray am not too sure. I would still go with ida, we will see when
it is relly removed.
If you need a bitfield why not use bitmap apis.
I dont like drivers implementing the basic logic which kernel provides
--
~Vinod