Hello,
> -----Original Message-----
> From: Jonas Gorski <jonas.gor...@gmail.com>
> Sent: Sunday, March 12, 2023 9:30 PM
> To: Mahapatra, Amit Kumar <amit.kumar-mahapa...@amd.com>
> Cc: broo...@kernel.org; miquel.ray...@bootlin.com; rich...@nod.at;
> vigne...@ti.com; ji...@kernel.org; tudor.amba...@microchip.com;
> praty...@kernel.org; Mehta, Sanju <sanju.me...@amd.com>; chin-
> ting_...@aspeedtech.com; c...@kaod.org; kdasu.k...@gmail.com;
> f.faine...@gmail.com; r...@broadcom.com; sbran...@broadcom.com;
> eaja...@linux.ibm.com; olte...@gmail.com; han...@nxp.com;
> john.ga...@huawei.com; shawn...@kernel.org; s.ha...@pengutronix.de;
> narmstr...@baylibre.com; khil...@baylibre.com;
> matthias....@gmail.com; haibo.c...@nxp.com; linus.wall...@linaro.org;
> dan...@zonque.org; haojian.zhu...@gmail.com; robert.jarz...@free.fr;
> agr...@kernel.org; bjorn.anders...@linaro.org; he...@sntech.de;
> krzysztof.kozlow...@linaro.org; a...@etezian.org;
> mcoquelin.st...@gmail.com; alexandre.tor...@foss.st.com;
> w...@csie.org; jernej.skra...@gmail.com; sam...@sholland.org;
> masahisa.koj...@linaro.org; jaswinder.si...@linaro.org;
> rost...@goodmis.org; mi...@redhat.com; l.stelm...@samsung.com;
> da...@davemloft.net; eduma...@google.com; k...@kernel.org;
> pab...@redhat.com; alex.ar...@gmail.com; ste...@datenfreihafen.org;
> kv...@kernel.org; james.schul...@cirrus.com; david.rho...@cirrus.com;
> tanur...@opensource.cirrus.com; r...@opensource.cirrus.com;
> pe...@perex.cz; ti...@suse.com; npig...@gmail.com;
> christophe.le...@csgroup.eu; m...@ellerman.id.au; o...@buserror.net;
> win...@126.com; yangyingli...@huawei.com;
> william.zh...@broadcom.com; kursad.o...@broadcom.com;
> anand.g...@broadcom.com; ra...@milecki.pl; git (AMD-Xilinx)
> <g...@amd.com>; linux-...@vger.kernel.org; linux-ker...@vger.kernel.org;
> j...@jms.id.au; and...@aj.id.au; radu_nicolae.pi...@upb.ro;
> nicolas.fe...@microchip.com; alexandre.bell...@bootlin.com;
> claudiu.bez...@microchip.com; bcm-kernel-feedback-l...@broadcom.com;
> fancer.lan...@gmail.com; ker...@pengutronix.de; feste...@gmail.com;
> linux-...@nxp.com; jbru...@baylibre.com;
> martin.blumensti...@googlemail.com; avifishma...@gmail.com;
> tmaimo...@gmail.com; tali.per...@gmail.com; vent...@google.com;
> yu...@google.com; benjaminf...@google.com; yogeshgaur...@gmail.com;
> konrad.dyb...@somainline.org; alim.akh...@samsung.com;
> ldewan...@nvidia.com; thierry.red...@gmail.com; jonath...@nvidia.com;
> Simek, Michal <michal.si...@amd.com>; linux-asp...@lists.ozlabs.org;
> open...@lists.ozlabs.org; linux-arm-ker...@lists.infradead.org; linux-rpi-
> ker...@lists.infradead.org; linux-amlo...@lists.infradead.org; linux-
> media...@lists.infradead.org; linux-arm-...@vger.kernel.org; linux-
> rockc...@lists.infradead.org; linux-samsung-...@vger.kernel.org; linux-
> st...@st-md-mailman.stormreply.com; linux-su...@lists.linux.dev; linux-
> te...@vger.kernel.org; net...@vger.kernel.org; linux-
> w...@vger.kernel.org; libertas-...@lists.infradead.org; linux-
> wirel...@vger.kernel.org; linux-...@lists.infradead.org; l...@metafoo.de;
> michael.henner...@analog.com; linux-...@vger.kernel.org;
> mich...@walle.cc; pal...@dabbelt.com; linux-ri...@lists.infradead.org;
> alsa-de...@alsa-project.org; patc...@opensource.cirrus.com; linuxppc-
> d...@lists.ozlabs.org; amitrkcian2...@gmail.com
> Subject: Re: [PATCH V6 09/15] spi: Add stacked and parallel memories
> support in SPI core
>
> Hi,
>
> On Fri, 10 Mar 2023 at 18:37, Amit Kumar Mahapatra <amit.kumar-
> mahapa...@amd.com> wrote:
> >
> > For supporting multiple CS the SPI device need to be aware of all the
> > CS values. So, the "chip_select" member in the spi_device structure is
> > now an array that holds all the CS values.
> >
> > spi_device structure now has a "cs_index_mask" member. This acts as an
> > index to the chip_select array. If nth bit of spi->cs_index_mask is
> > set then the driver would assert spi->chip_select[n].
> >
> > In parallel mode all the chip selects are asserted/de-asserted
> > simultaneously and each byte of data is stored in both devices, the
> > even bits in one, the odd bits in the other. The split is
> > automatically handled by the GQSPI controller. The GQSPI controller
> > supports a maximum of two flashes connected in parallel mode. A
> > "multi-cs-cap" flag is added in the spi controntroller data, through
> > ctlr->multi-cs-cap the spi core will make sure that the controller is
> > capable of handling multiple chip selects at once.
> >
> > For supporting multiple CS via GPIO the cs_gpiod member of the
> > spi_device structure is now an array that holds the gpio descriptor
> > for each chipselect.
> >
> > Multi CS support using GPIO is not tested due to unavailability of
> > necessary hardware setup.
>
> Can you pinmux your SPI controller's (cs) pins as GPIO? If so, you should be
> able use that for testing.
>
Xilinx Controller drivers that support multi cs are registered under
spi-mem framework. So even if I modify the pinmux the chip selection
will not go through the SPI core.
So, we cannot test the CS GPIO changes in SPI core on our platforms.
> >
> > Signed-off-by: Amit Kumar Mahapatra <amit.kumar-
> mahapa...@amd.com>
> > ---
> > drivers/spi/spi.c | 225 +++++++++++++++++++++++++++-------------
> > include/linux/spi/spi.h | 34 ++++--
> > 2 files changed, 182 insertions(+), 77 deletions(-)
> >
> > diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c index
> > c725b4bab7af..742bd688381c 100644
> > --- a/drivers/spi/spi.c
> > +++ b/drivers/spi/spi.c
> > @@ -612,10 +612,17 @@ static int spi_dev_check(struct device *dev,
> > void *data) {
> > struct spi_device *spi = to_spi_device(dev);
> > struct spi_device *new_spi = data;
> > -
> > - if (spi->controller == new_spi->controller &&
> > - spi_get_chipselect(spi, 0) == spi_get_chipselect(new_spi, 0))
> > - return -EBUSY;
> > + int idx, nw_idx;
> > +
> > + if (spi->controller == new_spi->controller) {
> > + for (idx = 0; idx < SPI_CS_CNT_MAX; idx++) {
> > + for (nw_idx = 0; nw_idx < SPI_CS_CNT_MAX; nw_idx++)
> > {
> > + if (spi_get_chipselect(spi, idx) ==
> > + spi_get_chipselect(new_spi, nw_idx))
> > + return -EBUSY;
> > + }
> > + }
>
> AFAICT unused chip selects are initialized to 0, so all single chip select
> devices
> would have it as their second one. This will then cause this check to reject
> every single chip select device after the first one. So you first need to make
> sure to only compare valid chip selects.
>
> So the loop condition should be something along idx <
> spi_get_num_chipselect(), not idx < SPI_CS_CNT_MAX.
>
Agreed, will update the loop condition as per num_cs.
> > + }
> > return 0;
> > }
> >
> > @@ -629,7 +636,7 @@ static int __spi_add_device(struct spi_device
> > *spi) {
> > struct spi_controller *ctlr = spi->controller;
> > struct device *dev = ctlr->dev.parent;
> > - int status;
> > + int status, idx;
> >
> > /*
> > * We need to make sure there's no other device with this @@
> > -638,8 +645,7 @@ static int __spi_add_device(struct spi_device *spi)
> > */
> > status = bus_for_each_dev(&spi_bus_type, NULL, spi, spi_dev_check);
> > if (status) {
> > - dev_err(dev, "chipselect %d already in use\n",
> > - spi_get_chipselect(spi, 0));
> > + dev_err(dev, "chipselect %d already in use\n",
> > + spi_get_chipselect(spi, 0));
>
> The message might be misleading for multi cs devices where the first one is
> free, but the second one is already in use.
>
> So maybe move this error message into spi_dev_check(), where you have
> that information available. You then even have the chance to state what is
> using the CS then, but that might be something for a different patch.
>
>
Agreed, I will move the error message to spi_dev_check().
> > return status;
> > }
> >
> > @@ -649,8 +655,10 @@ static int __spi_add_device(struct spi_device *spi)
> > return -ENODEV;
> > }
> >
> > - if (ctlr->cs_gpiods)
> > - spi_set_csgpiod(spi, 0,
> > ctlr->cs_gpiods[spi_get_chipselect(spi, 0)]);
> > + if (ctlr->cs_gpiods) {
> > + for (idx = 0; idx < SPI_CS_CNT_MAX; idx++)
> > + spi_set_csgpiod(spi, idx, ctlr-
> >cs_gpiods[spi_get_chipselect(spi, idx)]);
> > + }
> >
> > /*
> > * Drivers may modify this initial i/o setup, but will @@
> > -690,13 +698,15 @@ int spi_add_device(struct spi_device *spi) {
> > struct spi_controller *ctlr = spi->controller;
> > struct device *dev = ctlr->dev.parent;
> > - int status;
> > + int status, idx;
> >
> > - /* Chipselects are numbered 0..max; validate. */
> > - if (spi_get_chipselect(spi, 0) >= ctlr->num_chipselect) {
> > - dev_err(dev, "cs%d >= max %d\n", spi_get_chipselect(spi, 0),
> > - ctlr->num_chipselect);
> > - return -EINVAL;
> > + for (idx = 0; idx < SPI_CS_CNT_MAX; idx++) {
> > + /* Chipselects are numbered 0..max; validate. */
> > + if (spi_get_chipselect(spi, idx) >= ctlr->num_chipselect) {
> > + dev_err(dev, "cs%d >= max %d\n",
> > spi_get_chipselect(spi,
> idx),
> > + ctlr->num_chipselect);
> > + return -EINVAL;
> > + }
> > }
> >
> > /* Set the bus ID string */
> > @@ -713,12 +723,15 @@ static int spi_add_device_locked(struct
> > spi_device *spi) {
> > struct spi_controller *ctlr = spi->controller;
> > struct device *dev = ctlr->dev.parent;
> > + int idx;
> >
> > - /* Chipselects are numbered 0..max; validate. */
> > - if (spi_get_chipselect(spi, 0) >= ctlr->num_chipselect) {
> > - dev_err(dev, "cs%d >= max %d\n", spi_get_chipselect(spi, 0),
> > - ctlr->num_chipselect);
> > - return -EINVAL;
> > + for (idx = 0; idx < SPI_CS_CNT_MAX; idx++) {
> > + /* Chipselects are numbered 0..max; validate. */
> > + if (spi_get_chipselect(spi, idx) >= ctlr->num_chipselect) {
> > + dev_err(dev, "cs%d >= max %d\n",
> > spi_get_chipselect(spi,
> idx),
> > + ctlr->num_chipselect);
> > + return -EINVAL;
> > + }
> > }
> >
> > /* Set the bus ID string */
> > @@ -966,58 +979,119 @@ static void spi_res_release(struct
> > spi_controller *ctlr, struct spi_message *mes static void
> > spi_set_cs(struct spi_device *spi, bool enable, bool force) {
> > bool activate = enable;
> > + u32 cs_num = __ffs(spi->cs_index_mask);
> > + int idx;
> >
> > /*
> > - * Avoid calling into the driver (or doing delays) if the chip
> > select
> > - * isn't actually changing from the last time this was called.
> > + * In parallel mode all the chip selects are asserted/de-asserted
> > + * at once
> > */
> > - if (!force && ((enable && spi->controller->last_cs ==
> spi_get_chipselect(spi, 0)) ||
> > - (!enable && spi->controller->last_cs !=
> > spi_get_chipselect(spi,
> 0))) &&
> > - (spi->controller->last_cs_mode_high == (spi->mode &
> SPI_CS_HIGH)))
> > - return;
> > -
> > - trace_spi_set_cs(spi, activate);
> > -
> > - spi->controller->last_cs = enable ? spi_get_chipselect(spi, 0) : -1;
> > - spi->controller->last_cs_mode_high = spi->mode & SPI_CS_HIGH;
> > -
> > - if ((spi_get_csgpiod(spi, 0) || !spi->controller->set_cs_timing) &&
> !activate)
> > - spi_delay_exec(&spi->cs_hold, NULL);
> > -
> > - if (spi->mode & SPI_CS_HIGH)
> > - enable = !enable;
> > + if ((spi->cs_index_mask & SPI_PARALLEL_CS_MASK) ==
> SPI_PARALLEL_CS_MASK) {
> > + spi->controller->last_cs_mode_high = spi->mode &
> > + SPI_CS_HIGH;
> > +
> > + if ((spi_get_csgpiod(spi, 0) ||
> > !spi->controller->set_cs_timing) &&
> !activate)
> > + spi_delay_exec(&spi->cs_hold, NULL);
> > +
> > + if (spi->mode & SPI_CS_HIGH)
> > + enable = !enable;
> > +
> > + if (spi_get_csgpiod(spi, 0) && spi_get_csgpiod(spi, 1)) {
> > + if (!(spi->mode & SPI_NO_CS)) {
> > + /*
> > + * Historically ACPI has no means of the
> > GPIO polarity
> and
> > + * thus the SPISerialBus() resource defines
> > it on the per-
> chip
> > + * basis. In order to avoid a chain of
> > negations, the GPIO
> > + * polarity is considered being Active
> > High. Even for the
> cases
> > + * when _DSD() is involved (in the updated
> > versions of
> ACPI)
> > + * the GPIO CS polarity must be defined
> > Active High to
> avoid
> > + * ambiguity. That's why we use enable,
> > that takes
> SPI_CS_HIGH
> > + * into account.
> > + */
> > + if (has_acpi_companion(&spi->dev)) {
> > + for (idx = 0; idx < SPI_CS_CNT_MAX;
> > idx++)
> > +
> > gpiod_set_value_cansleep(spi_get_csgpiod(spi,
> idx),
> > +
> > !enable);
> > + } else {
> > + for (idx = 0; idx < SPI_CS_CNT_MAX;
> > idx++)
> > + /* Polarity handled by GPIO
> > library */
> > +
> > gpiod_set_value_cansleep(spi_get_csgpiod(spi,
> idx),
> > +
> > activate);
> > + }
> > + }
> > + /* Some SPI masters need both GPIO CS &
> > slave_select */
> > + if ((spi->controller->flags & SPI_MASTER_GPIO_SS) &&
> > + spi->controller->set_cs)
> > + spi->controller->set_cs(spi, !enable);
>
> > + else if (spi->controller->set_cs)
> > + spi->controller->set_cs(spi, !enable);
>
> this else if belongs to the following brace as the else of the if
> (spi_get_csgpiod(spi, 0) && spi_get_csgpiod(spi, 1). Currently it would make
Agreed, will fix it in the next series.
> the first check redundant, as the second case would always be true if the
> first
> one is.
>
> Actually shouldn't you iterate over the cs's here in case one is using
> set_cs() and the other one is gpiod? You can only get here if both are backed
> by gpiods. And you would only set the first cs, but not the second one. The -
> >set_cs() callback doesn't allow specifying which of the (multiple) cs's
> >should
> be set though.
>
After fixing the else if indentation we will get here if either one of the
CS support gpiod or both the CS support set_cs. Yes, one is using set_cs()
and the other one is gpiod use case handling is missing. I need to iterate
over the CS’s to find the CS GPIO, call gpiod_set_value_cansleep ( ) and
then call set_cs( ). In the set_cs( ) driver API the driver needs to first
check
if any of the cs_index_mask enabled CS's is not a CS GPIO and then enable
only the non-gpio CS.
Please let me your thoughts on this approach.
> > + }
> >
> > - if (spi_get_csgpiod(spi, 0)) {
> > - if (!(spi->mode & SPI_NO_CS)) {
> > - /*
> > - * Historically ACPI has no means of the GPIO
> > polarity and
> > - * thus the SPISerialBus() resource defines it on
> > the per-chip
> > - * basis. In order to avoid a chain of negations,
> > the GPIO
> > - * polarity is considered being Active High. Even
> > for the cases
> > - * when _DSD() is involved (in the updated versions
> > of ACPI)
> > - * the GPIO CS polarity must be defined Active High
> > to avoid
> > - * ambiguity. That's why we use enable, that takes
> SPI_CS_HIGH
> > - * into account.
> > - */
> > - if (has_acpi_companion(&spi->dev))
> > -
> > gpiod_set_value_cansleep(spi_get_csgpiod(spi, 0),
> !enable);
> > - else
> > - /* Polarity handled by GPIO library */
> > -
> > gpiod_set_value_cansleep(spi_get_csgpiod(spi, 0),
> activate);
> > + for (idx = 0; idx < SPI_CS_CNT_MAX; idx++) {
> > + if (spi_get_csgpiod(spi, idx) || !spi->controller-
> >set_cs_timing) {
> > + if (activate)
> > + spi_delay_exec(&spi->cs_setup,
> > NULL);
> > + else
> > + spi_delay_exec(&spi->cs_inactive,
> > NULL);
> > + }
>
> Won't you delay twice if both CS's are backed by gpiod (and the controller
> does not implement set_cs_timing)? You should probably break after the
> first or so.
>
True, I will add a check to avoid extra delay.
> I wonder if it would makes sense to have a helper function to set cs state to
> all cs's indicated by cs_index_mask so you can share most of the logic
> between the single and multi cs paths.
>
> Currently it seems both paths have a lot of code (and comment) duplication,
> with the difference being one path is touching one cs and the other two (or
> all).
>
Agreed, will update the logic.
> > }
> > - /* Some SPI masters need both GPIO CS & slave_select */
> > - if ((spi->controller->flags & SPI_MASTER_GPIO_SS) &&
> > - spi->controller->set_cs)
> > + } else {
> > + /*
> > + * Avoid calling into the driver (or doing delays) if the
> > chip select
> > + * isn't actually changing from the last time this was
> > called.
> > + */
> > + if (!force && ((enable && spi->controller->last_cs ==
> > + spi_get_chipselect(spi, cs_num)) ||
> > + (!enable && spi->controller->last_cs !=
> > + spi_get_chipselect(spi, cs_num))) &&
> > + (spi->controller->last_cs_mode_high ==
> > + (spi->mode & SPI_CS_HIGH)))
> > + return;
> > +
> > + trace_spi_set_cs(spi, activate);
> > +
> > + spi->controller->last_cs = enable ? spi_get_chipselect(spi,
> > cs_num)
> : -1;
> > + spi->controller->last_cs_mode_high = spi->mode &
> > + SPI_CS_HIGH;
> > +
> > + if ((spi_get_csgpiod(spi, cs_num) || !spi->controller-
> >set_cs_timing) && !activate)
> > + spi_delay_exec(&spi->cs_hold, NULL);
> > +
> > + if (spi->mode & SPI_CS_HIGH)
> > + enable = !enable;
> > +
> > + if (spi_get_csgpiod(spi, cs_num)) {
> > + if (!(spi->mode & SPI_NO_CS)) {
> > + /*
> > + * Historically ACPI has no means of the
> > GPIO polarity
> and
> > + * thus the SPISerialBus() resource defines
> > it on the per-
> chip
> > + * basis. In order to avoid a chain of
> > negations, the GPIO
> > + * polarity is considered being Active
> > High. Even for the
> cases
> > + * when _DSD() is involved (in the updated
> > versions of
> ACPI)
> > + * the GPIO CS polarity must be defined
> > Active High to
> avoid
> > + * ambiguity. That's why we use enable,
> > that takes
> SPI_CS_HIGH
> > + * into account.
> > + */
> > + if (has_acpi_companion(&spi->dev))
> > +
> > gpiod_set_value_cansleep(spi_get_csgpiod(spi,
> cs_num),
> > + !enable);
> > + else
> > + /* Polarity handled by GPIO library
> > */
> > +
> > gpiod_set_value_cansleep(spi_get_csgpiod(spi,
> cs_num),
> > + activate);
> > + }
> > + /* Some SPI masters need both GPIO CS &
> > slave_select */
> > + if ((spi->controller->flags & SPI_MASTER_GPIO_SS) &&
> > + spi->controller->set_cs)
> > + spi->controller->set_cs(spi, !enable);
> > + } else if (spi->controller->set_cs) {
> > spi->controller->set_cs(spi, !enable);
> > - } else if (spi->controller->set_cs) {
> > - spi->controller->set_cs(spi, !enable);
> > - }
> > + }
> >
> > - if (spi_get_csgpiod(spi, 0) || !spi->controller->set_cs_timing) {
> > - if (activate)
> > - spi_delay_exec(&spi->cs_setup, NULL);
> > - else
> > - spi_delay_exec(&spi->cs_inactive, NULL);
> > + if (spi_get_csgpiod(spi, cs_num) || !spi->controller-
> >set_cs_timing) {
> > + if (activate)
> > + spi_delay_exec(&spi->cs_setup, NULL);
> > + else
> > + spi_delay_exec(&spi->cs_inactive, NULL);
> > + }
> > }
> > }
> >
> > @@ -2246,8 +2320,8 @@ static void of_spi_parse_dt_cs_delay(struct
> > device_node *nc, static int of_spi_parse_dt(struct spi_controller *ctlr,
> struct spi_device *spi,
> > struct device_node *nc) {
> > - u32 value;
> > - int rc;
> > + u32 value, cs[SPI_CS_CNT_MAX] = {0};
> > + int rc, idx;
> >
> > /* Mode (clock phase/polarity/etc.) */
> > if (of_property_read_bool(nc, "spi-cpha")) @@ -2320,13
> > +2394,21 @@ static int of_spi_parse_dt(struct spi_controller *ctlr, struct
> spi_device *spi,
> > }
> >
> > /* Device address */
> > - rc = of_property_read_u32(nc, "reg", &value);
> > - if (rc) {
> > + rc = of_property_read_variable_u32_array(nc, "reg", &cs[0], 1,
> > + SPI_CS_CNT_MAX);
> > + if (rc < 0 || rc > ctlr->num_chipselect) {
> > dev_err(&ctlr->dev, "%pOF has no valid 'reg' property
> > (%d)\n",
> > nc, rc);
> > return rc;
> > + } else if ((of_property_read_bool(nc, "parallel-memories")) &&
> > + (!ctlr->multi_cs_cap)) {
> > + dev_err(&ctlr->dev, "SPI controller doesn't support multi
> > CS\n");
> > + return -EINVAL;
> > }
> > - spi_set_chipselect(spi, 0, value);
> > + for (idx = 0; idx < rc; idx++)
> > + spi_set_chipselect(spi, idx, cs[idx]);
> > + /* By default set the spi->cs_index_mask as 1 */
> > + spi->cs_index_mask = 0x01;
> >
> > /* Device speed */
> > if (!of_property_read_u32(nc, "spi-max-frequency", &value)) @@
> > -3846,6 +3928,7 @@ static int __spi_validate(struct spi_device *spi, struct
> spi_message *message)
> > struct spi_controller *ctlr = spi->controller;
> > struct spi_transfer *xfer;
> > int w_size;
> > + u32 cs_num = __ffs(spi->cs_index_mask);
> >
> > if (list_empty(&message->transfers))
> > return -EINVAL;
> > @@ -3858,7 +3941,7 @@ static int __spi_validate(struct spi_device *spi,
> struct spi_message *message)
> > * cs_change is set for each transfer.
> > */
> > if ((spi->mode & SPI_CS_WORD) && (!(ctlr->mode_bits &
> SPI_CS_WORD) ||
> > - spi_get_csgpiod(spi, 0))) {
> > + spi_get_csgpiod(spi,
> > + cs_num))) {
>
> Wouldn't you need to check for any of the cs_index_mask enabled CS's, and
> not just the first one? AFAICT you would currently fail to catch a
> SPI_CS_WORD transfer with both cs enabled where the first one is a
> SPI_CS_WORD capable native CS and the second one a gpiod.
>
That’s true, I will add a loop and check for each of the cs_index_mask
enabled CS's.
> > size_t maxsize;
> > int ret;
> >
> > diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h index
> > bdb35a91b4bf..452682aa1a39 100644
> > --- a/include/linux/spi/spi.h
> > +++ b/include/linux/spi/spi.h
> > @@ -19,6 +19,11 @@
> > #include <linux/acpi.h>
> > #include <linux/u64_stats_sync.h>
> >
> > +/* Max no. of CS supported per spi device */ #define SPI_CS_CNT_MAX 2
> > +
> > +/* chip select mask */
> > +#define SPI_PARALLEL_CS_MASK (BIT(0) | BIT(1))
> > struct dma_chan;
> > struct software_node;
> > struct ptp_system_timestamp;
> > @@ -166,6 +171,7 @@ extern void
> spi_transfer_cs_change_delay_exec(struct spi_message *msg,
> > * deasserted. If @cs_change_delay is used from @spi_transfer, then
> the
> > * two delays will be added up.
> > * @pcpu_statistics: statistics for the spi_device
> > + * @cs_index_mask: Bit mask of the active chipselect(s) in the
> > + chipselect array
> > *
> > * A @spi_device is used to interchange data between an SPI slave
> > * (usually a discrete chip) and CPU memory.
> > @@ -181,7 +187,7 @@ struct spi_device {
> > struct spi_controller *controller;
> > struct spi_controller *master; /* Compatibility layer */
> > u32 max_speed_hz;
> > - u8 chip_select;
> > + u8 chip_select[SPI_CS_CNT_MAX];
> > u8 bits_per_word;
> > bool rt;
> > #define SPI_NO_TX BIT(31) /* No transmit wire */
> > @@ -202,7 +208,7 @@ struct spi_device {
> > void *controller_data;
> > char modalias[SPI_NAME_SIZE];
> > const char *driver_override;
> > - struct gpio_desc *cs_gpiod; /* Chip select gpio desc */
> > + struct gpio_desc *cs_gpiod[SPI_CS_CNT_MAX]; /* Chip
> > select
> gpio desc */
> > struct spi_delay word_delay; /* Inter-word delay */
> > /* CS delays */
> > struct spi_delay cs_setup;
> > @@ -212,6 +218,13 @@ struct spi_device {
> > /* The statistics */
> > struct spi_statistics __percpu *pcpu_statistics;
> >
> > + /* Bit mask of the chipselect(s) that the driver need to use from
> > + * the chipselect array.When the controller is capable to handle
> > + * multiple chip selects & memories are connected in parallel
> > + * then more than one bit need to be set in cs_index_mask.
> > + */
> > + u32 cs_index_mask : 2;
>
> SPI_CS_CNT_MAX?
>
Agreed, will replace 2 with SPI_CS_CNT_MAX.
> > +
> > /*
> > * likely need more hooks for more protocol options affecting how
> > * the controller talks to each chip, like:
> > @@ -268,22 +281,22 @@ static inline void *spi_get_drvdata(struct
> > spi_device *spi)
> >
> > static inline u8 spi_get_chipselect(struct spi_device *spi, u8 idx)
> > {
> > - return spi->chip_select;
> > + return spi->chip_select[idx];
> > }
> >
> > static inline void spi_set_chipselect(struct spi_device *spi, u8 idx,
> > u8 chipselect) {
> > - spi->chip_select = chipselect;
> > + spi->chip_select[idx] = chipselect;
> > }
> >
> > static inline struct gpio_desc *spi_get_csgpiod(struct spi_device
> > *spi, u8 idx) {
> > - return spi->cs_gpiod;
> > + return spi->cs_gpiod[idx];
> > }
> >
> > static inline void spi_set_csgpiod(struct spi_device *spi, u8 idx,
> > struct gpio_desc *csgpiod) {
> > - spi->cs_gpiod = csgpiod;
> > + spi->cs_gpiod[idx] = csgpiod;
> > }
> >
> > /**
> > @@ -388,6 +401,8 @@ extern struct spi_device
> *spi_new_ancillary_device(struct spi_device *spi, u8 ch
> > * @bus_lock_spinlock: spinlock for SPI bus locking
> > * @bus_lock_mutex: mutex for exclusion of multiple callers
> > * @bus_lock_flag: indicates that the SPI bus is locked for exclusive
> > use
> > + * @multi_cs_cap: indicates that the SPI Controller can assert/de-assert
> > + * more than one chip select at once.
> > * @setup: updates the device mode and clocking records used by a
> > * device's SPI controller; protocol code may call this. This
> > * must fail if an unrecognized or unsupported mode is requested.
> > @@ -585,6 +600,13 @@ struct spi_controller {
> > /* Flag indicating that the SPI bus is locked for exclusive use */
> > bool bus_lock_flag;
> >
> > + /*
> > + * Flag indicating that the spi-controller has multi chip select
> > + * capability and can assert/de-assert more than one chip select
> > + * at once.
> > + */
> > + bool multi_cs_cap;
>
> I admit I haven't followed the first iterations, but Is there a reason this
> isn't a
> SPI_XXX flag in spi.h? There seem to be quite a few free bits left.
>
Yes, it can be a SPI_XX flag as well. I will add a flag & remove this
structure member.
> I would think multi_cs can be emulated (somewhat) via gpiod for the second
> CS as long as the controller supports set_cs() (and SPI_NO_CS?).
>
It is not just about handling the CS's, but rather this flag indicates
that the controller can communicate (reading & writing data) with
both the devices simultaneously.
Regards,
Amit
> > +
> > /* Setup mode and clock, etc (spi driver may call many times).
> > *
> > * IMPORTANT: this may be called when transfers to another
> > --
> > 2.25.1
> >
>
> Regards
> Jonas
