Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 27 Jan 2019 19:04:02 +0000 (UTC)
From:      Marius Strobl <marius@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-12@freebsd.org
Subject:   svn commit: r343503 - in stable/12/sys: arm/broadcom/bcm2835 arm/nvidia arm/ti dev/sdhci
Message-ID:  <201901271904.x0RJ4264085105@repo.freebsd.org>

next in thread | raw e-mail | index | archive | help
Author: marius
Date: Sun Jan 27 19:04:02 2019
New Revision: 343503
URL: https://svnweb.freebsd.org/changeset/base/343503

Log:
  MFC: r342634
  
  o Don't allocate resources for SDMA in sdhci(4) if the controller or the
    front-end doesn't support SDMA or the latter implements a platform-
    specific transfer method instead. While at it, factor out allocation
    and freeing of SDMA resources to sdhci_dma_{alloc,free}() in order to
    keep the code more readable when adding support for ADMA variants.
  
  o Base the size of the SDMA bounce buffer on MAXPHYS up to the maximum
    of 512 KiB instead of using a fixed 4-KiB-buffer. With the default
    MAXPHYS of 128 KiB and depending on the controller and medium, this
    reduces the number of SDHCI interrupts by a factor of ~16 to ~32 on
    sequential reads while an increase of throughput of up to ~84 % was
    seen.
  
    Front-ends for broken controllers that only support an SDMA buffer
    boundary of a specific size may set SDHCI_QUIRK_BROKEN_SDMA_BOUNDARY
    and supply a size via struct sdhci_slot. According to Linux, only
    Qualcomm MSM-type SDHCI controllers are affected by this, though.
  
    Requested by: Shreyank Amartya (unconditional bump to 512 KiB)
  
  o Introduce a SDHCI_DEPEND macro for specifying the dependency of the
    front-end modules on the sdhci(4) one and bump the module version
    of sdhci(4) to 2 via an also newly introduced SDHCI_VERSION in order
    to ensure that all components are in sync WRT struct sdhci_slot.
  
  o In sdhci(4):
    - Make pointers const were applicable,
    - replace a few device_printf(9) calls with slot_printf() for
      consistency, and
    - sync some local functions with their prototypes WRT static.

Modified:
  stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhci.c
  stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhost.c
  stable/12/sys/arm/nvidia/tegra_sdhci.c
  stable/12/sys/arm/ti/ti_sdhci.c
  stable/12/sys/dev/sdhci/fsl_sdhci.c
  stable/12/sys/dev/sdhci/sdhci.c
  stable/12/sys/dev/sdhci/sdhci.h
  stable/12/sys/dev/sdhci/sdhci_acpi.c
  stable/12/sys/dev/sdhci/sdhci_fdt.c
  stable/12/sys/dev/sdhci/sdhci_pci.c
  stable/12/sys/dev/sdhci/sdhci_xenon.c
Directory Properties:
  stable/12/   (props changed)

Modified: stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhci.c
==============================================================================
--- stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhci.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhci.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -683,7 +683,7 @@ static driver_t bcm_sdhci_driver = {
 
 DRIVER_MODULE(sdhci_bcm, simplebus, bcm_sdhci_driver, bcm_sdhci_devclass,
     NULL, NULL);
-MODULE_DEPEND(sdhci_bcm, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_bcm);
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_bcm);
 #endif

Modified: stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhost.c
==============================================================================
--- stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhost.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/arm/broadcom/bcm2835/bcm2835_sdhost.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -1287,7 +1287,7 @@ static driver_t bcm_sdhost_driver = {
 
 DRIVER_MODULE(sdhost_bcm, simplebus, bcm_sdhost_driver, bcm_sdhost_devclass,
     NULL, NULL);
-MODULE_DEPEND(sdhost_bcm, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhost_bcm);
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhost_bcm);
 #endif

Modified: stable/12/sys/arm/nvidia/tegra_sdhci.c
==============================================================================
--- stable/12/sys/arm/nvidia/tegra_sdhci.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/arm/nvidia/tegra_sdhci.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -465,7 +465,7 @@ static DEFINE_CLASS_0(sdhci, tegra_sdhci_driver, tegra
     sizeof(struct tegra_sdhci_softc));
 DRIVER_MODULE(sdhci_tegra, simplebus, tegra_sdhci_driver, tegra_sdhci_devclass,
     NULL, NULL);
-MODULE_DEPEND(sdhci_tegra, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_tegra);
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci);
 #endif

Modified: stable/12/sys/arm/ti/ti_sdhci.c
==============================================================================
--- stable/12/sys/arm/ti/ti_sdhci.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/arm/ti/ti_sdhci.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -755,7 +755,7 @@ static driver_t ti_sdhci_driver = {
 
 DRIVER_MODULE(sdhci_ti, simplebus, ti_sdhci_driver, ti_sdhci_devclass, NULL,
     NULL);
-MODULE_DEPEND(sdhci_ti, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_ti);
 
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_ti);

Modified: stable/12/sys/dev/sdhci/fsl_sdhci.c
==============================================================================
--- stable/12/sys/dev/sdhci/fsl_sdhci.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/fsl_sdhci.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -1005,7 +1005,7 @@ static driver_t fsl_sdhci_driver = {
 
 DRIVER_MODULE(sdhci_fsl, simplebus, fsl_sdhci_driver, fsl_sdhci_devclass,
     NULL, NULL);
-MODULE_DEPEND(sdhci_fsl, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_fsl);
 
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_fsl);

Modified: stable/12/sys/dev/sdhci/sdhci.c
==============================================================================
--- stable/12/sys/dev/sdhci/sdhci.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/sdhci.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -36,6 +36,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/conf.h>
 #include <sys/kernel.h>
 #include <sys/kobj.h>
+#include <sys/libkern.h>
 #include <sys/lock.h>
 #include <sys/malloc.h>
 #include <sys/module.h>
@@ -108,19 +109,20 @@ static void sdhci_retune(void *arg);
 static void sdhci_set_clock(struct sdhci_slot *slot, uint32_t clock);
 static void sdhci_set_power(struct sdhci_slot *slot, u_char power);
 static void sdhci_set_transfer_mode(struct sdhci_slot *slot,
-   struct mmc_data *data);
+   const struct mmc_data *data);
 static void sdhci_start(struct sdhci_slot *slot);
 static void sdhci_timeout(void *arg);
 static void sdhci_start_command(struct sdhci_slot *slot,
    struct mmc_command *cmd);
-static void sdhci_start_data(struct sdhci_slot *slot, struct mmc_data *data);
+static void sdhci_start_data(struct sdhci_slot *slot,
+   const struct mmc_data *data);
 static void sdhci_write_block_pio(struct sdhci_slot *slot);
 static void sdhci_transfer_pio(struct sdhci_slot *slot);
 
 #ifdef MMCCAM
 /* CAM-related */
 static void sdhci_cam_action(struct cam_sim *sim, union ccb *ccb);
-static int sdhci_cam_get_possible_host_clock(struct sdhci_slot *slot,
+static int sdhci_cam_get_possible_host_clock(const struct sdhci_slot *slot,
     int proposed_clock);
 static void sdhci_cam_handle_mmcio(struct cam_sim *sim, union ccb *ccb);
 static void sdhci_cam_poll(struct cam_sim *sim);
@@ -130,12 +132,14 @@ static int sdhci_cam_update_ios(struct sdhci_slot *slo
 #endif
 
 /* helper routines */
+static int sdhci_dma_alloc(struct sdhci_slot *slot);
+static void sdhci_dma_free(struct sdhci_slot *slot);
 static void sdhci_dumpregs(struct sdhci_slot *slot);
 static void sdhci_getaddr(void *arg, bus_dma_segment_t *segs, int nsegs,
     int error);
-static int slot_printf(struct sdhci_slot *slot, const char * fmt, ...)
+static int slot_printf(const struct sdhci_slot *slot, const char * fmt, ...)
     __printflike(2, 3);
-static uint32_t sdhci_tuning_intmask(struct sdhci_slot *slot);
+static uint32_t sdhci_tuning_intmask(const struct sdhci_slot *slot);
 
 #define	SDHCI_LOCK(_slot)		mtx_lock(&(_slot)->mtx)
 #define	SDHCI_UNLOCK(_slot)		mtx_unlock(&(_slot)->mtx)
@@ -179,7 +183,7 @@ sdhci_getaddr(void *arg, bus_dma_segment_t *segs, int 
 }
 
 static int
-slot_printf(struct sdhci_slot *slot, const char * fmt, ...)
+slot_printf(const struct sdhci_slot *slot, const char * fmt, ...)
 {
 	va_list ap;
 	int retval;
@@ -290,7 +294,7 @@ sdhci_reset(struct sdhci_slot *slot, uint8_t mask)
 }
 
 static uint32_t
-sdhci_tuning_intmask(struct sdhci_slot *slot)
+sdhci_tuning_intmask(const struct sdhci_slot *slot)
 {
 	uint32_t intmask;
 
@@ -737,55 +741,94 @@ sdhci_card_poll(void *arg)
 	    sdhci_card_poll, slot);
 }
 
-int
-sdhci_init_slot(device_t dev, struct sdhci_slot *slot, int num)
+static int
+sdhci_dma_alloc(struct sdhci_slot *slot)
 {
-	kobjop_desc_t kobj_desc;
-	kobj_method_t *kobj_method;
-	uint32_t caps, caps2, freq, host_caps;
 	int err;
 
-	SDHCI_LOCK_INIT(slot);
+	if (!(slot->quirks & SDHCI_QUIRK_BROKEN_SDMA_BOUNDARY)) {
+		if (MAXPHYS <= 1024 * 4)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_4K;
+		else if (MAXPHYS <= 1024 * 8)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_8K;
+		else if (MAXPHYS <= 1024 * 16)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_16K;
+		else if (MAXPHYS <= 1024 * 32)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_32K;
+		else if (MAXPHYS <= 1024 * 64)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_64K;
+		else if (MAXPHYS <= 1024 * 128)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_128K;
+		else if (MAXPHYS <= 1024 * 256)
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_256K;
+		else
+			slot->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_512K;
+	}
+	slot->sdma_bbufsz = SDHCI_SDMA_BNDRY_TO_BBUFSZ(slot->sdma_boundary);
 
-	slot->num = num;
-	slot->bus = dev;
-
-	/* Allocate DMA tag. */
-	err = bus_dma_tag_create(bus_get_dma_tag(dev),
-	    DMA_BLOCK_SIZE, 0, BUS_SPACE_MAXADDR_32BIT,
-	    BUS_SPACE_MAXADDR, NULL, NULL,
-	    DMA_BLOCK_SIZE, 1, DMA_BLOCK_SIZE,
-	    BUS_DMA_ALLOCNOW, NULL, NULL,
-	    &slot->dmatag);
+	/*
+	 * Allocate the DMA tag for an SDMA bounce buffer.
+	 * Note that the SDHCI specification doesn't state any alignment
+	 * constraint for the SDMA system address.  However, controllers
+	 * typically ignore the SDMA boundary bits in SDHCI_DMA_ADDRESS when
+	 * forming the actual address of data, requiring the SDMA buffer to
+	 * be aligned to the SDMA boundary.
+	 */
+	err = bus_dma_tag_create(bus_get_dma_tag(slot->bus), slot->sdma_bbufsz,
+	    0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
+	    slot->sdma_bbufsz, 1, slot->sdma_bbufsz, BUS_DMA_ALLOCNOW,
+	    NULL, NULL, &slot->dmatag);
 	if (err != 0) {
-		device_printf(dev, "Can't create DMA tag\n");
-		SDHCI_LOCK_DESTROY(slot);
+		slot_printf(slot, "Can't create DMA tag for SDMA\n");
 		return (err);
 	}
-	/* Allocate DMA memory. */
+	/* Allocate DMA memory for the SDMA bounce buffer. */
 	err = bus_dmamem_alloc(slot->dmatag, (void **)&slot->dmamem,
 	    BUS_DMA_NOWAIT, &slot->dmamap);
 	if (err != 0) {
-		device_printf(dev, "Can't alloc DMA memory\n");
+		slot_printf(slot, "Can't alloc DMA memory for SDMA\n");
 		bus_dma_tag_destroy(slot->dmatag);
-		SDHCI_LOCK_DESTROY(slot);
 		return (err);
 	}
-	/* Map the memory. */
+	/* Map the memory of the SDMA bounce buffer. */
 	err = bus_dmamap_load(slot->dmatag, slot->dmamap,
-	    (void *)slot->dmamem, DMA_BLOCK_SIZE,
-	    sdhci_getaddr, &slot->paddr, 0);
+	    (void *)slot->dmamem, slot->sdma_bbufsz, sdhci_getaddr,
+	    &slot->paddr, 0);
 	if (err != 0 || slot->paddr == 0) {
-		device_printf(dev, "Can't load DMA memory\n");
+		slot_printf(slot, "Can't load DMA memory for SDMA\n");
 		bus_dmamem_free(slot->dmatag, slot->dmamem, slot->dmamap);
 		bus_dma_tag_destroy(slot->dmatag);
-		SDHCI_LOCK_DESTROY(slot);
 		if (err)
 			return (err);
 		else
 			return (EFAULT);
 	}
 
+	return (0);
+}
+
+static void
+sdhci_dma_free(struct sdhci_slot *slot)
+{
+
+	bus_dmamap_unload(slot->dmatag, slot->dmamap);
+	bus_dmamem_free(slot->dmatag, slot->dmamem, slot->dmamap);
+	bus_dma_tag_destroy(slot->dmatag);
+}
+
+int
+sdhci_init_slot(device_t dev, struct sdhci_slot *slot, int num)
+{
+	kobjop_desc_t kobj_desc;
+	kobj_method_t *kobj_method;
+	uint32_t caps, caps2, freq, host_caps;
+	int err;
+
+	SDHCI_LOCK_INIT(slot);
+
+	slot->num = num;
+	slot->bus = dev;
+
 	slot->version = (RD2(slot, SDHCI_HOST_VERSION)
 		>> SDHCI_SPEC_VER_SHIFT) & SDHCI_SPEC_VER_MASK;
 	if (slot->quirks & SDHCI_QUIRK_MISSING_CAPS) {
@@ -801,12 +844,8 @@ sdhci_init_slot(device_t dev, struct sdhci_slot *slot,
 	if (slot->version >= SDHCI_SPEC_300) {
 		if ((caps & SDHCI_SLOTTYPE_MASK) != SDHCI_SLOTTYPE_REMOVABLE &&
 		    (caps & SDHCI_SLOTTYPE_MASK) != SDHCI_SLOTTYPE_EMBEDDED) {
-			device_printf(dev,
+			slot_printf(slot,
 			    "Driver doesn't support shared bus slots\n");
-			bus_dmamap_unload(slot->dmatag, slot->dmamap);
-			bus_dmamem_free(slot->dmatag, slot->dmamem,
-			    slot->dmamap);
-			bus_dma_tag_destroy(slot->dmatag);
 			SDHCI_LOCK_DESTROY(slot);
 			return (ENXIO);
 		} else if ((caps & SDHCI_SLOTTYPE_MASK) ==
@@ -830,7 +869,7 @@ sdhci_init_slot(device_t dev, struct sdhci_slot *slot,
 	 */
 	if (slot->max_clk == 0) {
 		slot->max_clk = SDHCI_DEFAULT_MAX_FREQ * 1000000;
-		device_printf(dev, "Hardware doesn't specify base clock "
+		slot_printf(slot, "Hardware doesn't specify base clock "
 		    "frequency, using %dMHz as default.\n",
 		    SDHCI_DEFAULT_MAX_FREQ);
 	}
@@ -851,7 +890,7 @@ sdhci_init_slot(device_t dev, struct sdhci_slot *slot,
 	 * max timeout, but still mention it.
 	 */
 	if (slot->timeout_clk == 0) {
-		device_printf(dev, "Hardware doesn't specify timeout clock "
+		slot_printf(slot, "Hardware doesn't specify timeout clock "
 		    "frequency, setting BROKEN_TIMEOUT quirk.\n");
 		slot->quirks |= SDHCI_QUIRK_BROKEN_TIMEOUT_VAL;
 	}
@@ -867,7 +906,7 @@ sdhci_init_slot(device_t dev, struct sdhci_slot *slot,
 	if ((caps & SDHCI_CAN_VDD_180) && (slot->opt & SDHCI_SLOT_EMBEDDED))
 	    slot->host.host_ocr |= MMC_OCR_LOW_VOLTAGE;
 	if (slot->host.host_ocr == 0) {
-		device_printf(dev, "Hardware doesn't report any "
+		slot_printf(slot, "Hardware doesn't report any "
 		    "support voltages.\n");
 	}
 
@@ -953,7 +992,7 @@ no_tuning:
 			slot->retune_count = (caps2 & SDHCI_RETUNE_CNT_MASK) >>
 			    SDHCI_RETUNE_CNT_SHIFT;
 			if (slot->retune_count > 0xb) {
-				device_printf(dev, "Unknown re-tuning count "
+				slot_printf(slot, "Unknown re-tuning count "
 				    "%x, using 1 sec\n", slot->retune_count);
 				slot->retune_count = 1;
 			} else if (slot->retune_count != 0)
@@ -1012,6 +1051,19 @@ no_tuning:
 	if (slot->opt & SDHCI_PLATFORM_TRANSFER)
 		slot->opt &= ~SDHCI_HAVE_DMA;
 
+	if (slot->opt & SDHCI_HAVE_DMA) {
+		err = sdhci_dma_alloc(slot);
+		if (err != 0) {
+			if (slot->opt & SDHCI_TUNING_SUPPORTED) {
+				free(slot->tune_req, M_DEVBUF);
+				free(slot->tune_cmd, M_DEVBUF);
+				free(slot->tune_data, M_DEVBUF);
+			}
+			SDHCI_LOCK_DESTROY(slot);
+			return (err);
+		}
+	}
+
 	if (bootverbose || sdhci_debug) {
 		slot_printf(slot,
 		    "%uMHz%s %s VDD:%s%s%s VCCQ: 3.3V%s%s DRV: B%s%s%s %s %s\n",
@@ -1109,9 +1161,8 @@ sdhci_cleanup_slot(struct sdhci_slot *slot)
 	SDHCI_LOCK(slot);
 	sdhci_reset(slot, SDHCI_RESET_ALL);
 	SDHCI_UNLOCK(slot);
-	bus_dmamap_unload(slot->dmatag, slot->dmamap);
-	bus_dmamem_free(slot->dmatag, slot->dmamem, slot->dmamap);
-	bus_dma_tag_destroy(slot->dmatag);
+	if (slot->opt & SDHCI_HAVE_DMA)
+		sdhci_dma_free(slot);
 	if (slot->opt & SDHCI_TUNING_SUPPORTED) {
 		free(slot->tune_req, M_DEVBUF);
 		free(slot->tune_cmd, M_DEVBUF);
@@ -1175,7 +1226,7 @@ sdhci_generic_get_card_present(device_t brdev __unused
 void
 sdhci_generic_set_uhs_timing(device_t brdev __unused, struct sdhci_slot *slot)
 {
-	struct mmc_ios *ios;
+	const struct mmc_ios *ios;
 	uint16_t hostctrl2;
 
 	if (slot->version < SDHCI_SPEC_300)
@@ -1308,7 +1359,7 @@ int
 sdhci_generic_tune(device_t brdev __unused, device_t reqdev, bool hs400)
 {
 	struct sdhci_slot *slot = device_get_ivars(reqdev);
-	struct mmc_ios *ios = &slot->host.ios;
+	const struct mmc_ios *ios = &slot->host.ios;
 	struct mmc_command *tune_cmd;
 	struct mmc_data *tune_data;
 	uint32_t opcode;
@@ -1564,7 +1615,7 @@ sdhci_timeout(void *arg)
 }
 
 static void
-sdhci_set_transfer_mode(struct sdhci_slot *slot, struct mmc_data *data)
+sdhci_set_transfer_mode(struct sdhci_slot *slot, const struct mmc_data *data)
 {
 	uint16_t mode;
 
@@ -1766,9 +1817,9 @@ sdhci_finish_command(struct sdhci_slot *slot)
 }
 
 static void
-sdhci_start_data(struct sdhci_slot *slot, struct mmc_data *data)
+sdhci_start_data(struct sdhci_slot *slot, const struct mmc_data *data)
 {
-	uint32_t target_timeout, current_timeout;
+	uint32_t blkcnt, blksz, current_timeout, sdma_bbufsz, target_timeout;
 	uint8_t div;
 
 	if (data == NULL && (slot->curcmd->flags & MMC_RSP_BUSY) == 0) {
@@ -1804,7 +1855,7 @@ sdhci_start_data(struct sdhci_slot *slot, struct mmc_d
 	/* Use DMA if possible. */
 	if ((slot->opt & SDHCI_HAVE_DMA))
 		slot->flags |= SDHCI_USE_DMA;
-	/* If data is small, broken DMA may return zeroes instead of data, */
+	/* If data is small, broken DMA may return zeroes instead of data. */
 	if ((slot->quirks & SDHCI_QUIRK_BROKEN_TIMINGS) &&
 	    (data->len <= 512))
 		slot->flags &= ~SDHCI_USE_DMA;
@@ -1814,20 +1865,22 @@ sdhci_start_data(struct sdhci_slot *slot, struct mmc_d
 		slot->flags &= ~SDHCI_USE_DMA;
 	/* Load DMA buffer. */
 	if (slot->flags & SDHCI_USE_DMA) {
+		sdma_bbufsz = slot->sdma_bbufsz;
 		if (data->flags & MMC_DATA_READ)
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_PREREAD);
 		else {
-			memcpy(slot->dmamem, data->data,
-			    (data->len < DMA_BLOCK_SIZE) ?
-			    data->len : DMA_BLOCK_SIZE);
+			memcpy(slot->dmamem, data->data, ulmin(data->len,
+			    sdma_bbufsz));
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_PREWRITE);
 		}
 		WR4(slot, SDHCI_DMA_ADDRESS, slot->paddr);
-		/* Interrupt aggregation: Mask border interrupt
-		 * for the last page and unmask else. */
-		if (data->len == DMA_BLOCK_SIZE)
+		/*
+		 * Interrupt aggregation: Mask border interrupt for the last
+		 * bounce buffer and unmask otherwise.
+		 */
+		if (data->len == sdma_bbufsz)
 			slot->intmask &= ~SDHCI_INT_DMA_END;
 		else
 			slot->intmask |= SDHCI_INT_DMA_END;
@@ -1835,16 +1888,15 @@ sdhci_start_data(struct sdhci_slot *slot, struct mmc_d
 	}
 	/* Current data offset for both PIO and DMA. */
 	slot->offset = 0;
-	/* Set block size and request IRQ on 4K border. */
-	WR2(slot, SDHCI_BLOCK_SIZE, SDHCI_MAKE_BLKSZ(DMA_BOUNDARY,
-	    (data->len < 512) ? data->len : 512));
+	/* Set block size and request border interrupts on the SDMA boundary. */
+	blksz = SDHCI_MAKE_BLKSZ(slot->sdma_boundary, ulmin(data->len, 512));
+	WR2(slot, SDHCI_BLOCK_SIZE, blksz);
 	/* Set block count. */
-	WR2(slot, SDHCI_BLOCK_COUNT, (data->len + 511) / 512);
-
+	blkcnt = howmany(data->len, 512);
+	WR2(slot, SDHCI_BLOCK_COUNT, blkcnt);
 	if (__predict_false(sdhci_debug > 1))
-		slot_printf(slot, "Block size: %02x, count %lu\n",
-		    (unsigned int)SDHCI_MAKE_BLKSZ(DMA_BOUNDARY, (data->len < 512) ? data->len : 512),
-		    (unsigned long)(data->len + 511) / 512);
+		slot_printf(slot, "Blk size: 0x%08x | Blk cnt:  0x%08x\n",
+		    blksz, blkcnt);
 }
 
 void
@@ -1868,7 +1920,7 @@ sdhci_finish_data(struct sdhci_slot *slot)
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_POSTREAD);
 			memcpy((u_char*)data->data + slot->offset, slot->dmamem,
-			    (left < DMA_BLOCK_SIZE) ? left : DMA_BLOCK_SIZE);
+			    ulmin(left, slot->sdma_bbufsz));
 		} else
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_POSTWRITE);
@@ -1892,15 +1944,14 @@ sdhci_finish_data(struct sdhci_slot *slot)
 static void
 sdhci_start(struct sdhci_slot *slot)
 {
-        union ccb *ccb;
+	union ccb *ccb;
+	struct ccb_mmcio *mmcio;
 
 	ccb = slot->ccb;
 	if (ccb == NULL)
 		return;
 
-        struct ccb_mmcio *mmcio;
 	mmcio = &ccb->mmcio;
-
 	if (!(slot->flags & CMD_STARTED)) {
 		slot->flags |= CMD_STARTED;
 		sdhci_start_command(slot, &mmcio->cmd);
@@ -1932,7 +1983,7 @@ sdhci_start(struct sdhci_slot *slot)
 static void
 sdhci_start(struct sdhci_slot *slot)
 {
-	struct mmc_request *req;
+	const struct mmc_request *req;
 
 	req = slot->req;
 	if (req == NULL)
@@ -2061,6 +2112,7 @@ sdhci_data_irq(struct sdhci_slot *slot, uint32_t intma
 {
 	struct mmc_data *data;
 	size_t left;
+	uint32_t sdma_bbufsz;
 
 	if (!slot->curcmd) {
 		slot_printf(slot, "Got data interrupt 0x%08x, but "
@@ -2115,6 +2167,7 @@ sdhci_data_irq(struct sdhci_slot *slot, uint32_t intma
 	/* Handle DMA border. */
 	if (intmask & SDHCI_INT_DMA_END) {
 		data = slot->curcmd->data;
+		sdma_bbufsz = slot->sdma_bbufsz;
 
 		/* Unload DMA buffer ... */
 		left = data->len - slot->offset;
@@ -2122,26 +2175,28 @@ sdhci_data_irq(struct sdhci_slot *slot, uint32_t intma
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_POSTREAD);
 			memcpy((u_char*)data->data + slot->offset, slot->dmamem,
-			    (left < DMA_BLOCK_SIZE) ? left : DMA_BLOCK_SIZE);
+			    ulmin(left, sdma_bbufsz));
 		} else {
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_POSTWRITE);
 		}
 		/* ... and reload it again. */
-		slot->offset += DMA_BLOCK_SIZE;
+		slot->offset += sdma_bbufsz;
 		left = data->len - slot->offset;
 		if (data->flags & MMC_DATA_READ) {
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_PREREAD);
 		} else {
 			memcpy(slot->dmamem, (u_char*)data->data + slot->offset,
-			    (left < DMA_BLOCK_SIZE)? left : DMA_BLOCK_SIZE);
+			    ulmin(left, sdma_bbufsz));
 			bus_dmamap_sync(slot->dmatag, slot->dmamap,
 			    BUS_DMASYNC_PREWRITE);
 		}
-		/* Interrupt aggregation: Mask border interrupt
-		 * for the last page. */
-		if (left == DMA_BLOCK_SIZE) {
+		/*
+		 * Interrupt aggregation: Mask border interrupt for the last
+		 * bounce buffer.
+		 */
+		if (left == sdma_bbufsz) {
 			slot->intmask &= ~SDHCI_INT_DMA_END;
 			WR4(slot, SDHCI_SIGNAL_ENABLE, slot->intmask);
 		}
@@ -2264,7 +2319,7 @@ int
 sdhci_generic_read_ivar(device_t bus, device_t child, int which,
     uintptr_t *result)
 {
-	struct sdhci_slot *slot = device_get_ivars(child);
+	const struct sdhci_slot *slot = device_get_ivars(child);
 
 	switch (which) {
 	default:
@@ -2427,6 +2482,7 @@ sdhci_generic_write_ivar(device_t bus, device_t child,
 void
 sdhci_start_slot(struct sdhci_slot *slot)
 {
+
         if ((slot->devq = cam_simq_alloc(1)) == NULL) {
                 goto fail;
         }
@@ -2586,7 +2642,8 @@ sdhci_cam_poll(struct cam_sim *sim)
 }
 
 static int
-sdhci_cam_get_possible_host_clock(struct sdhci_slot *slot, int proposed_clock)
+sdhci_cam_get_possible_host_clock(const struct sdhci_slot *slot,
+    int proposed_clock)
 {
 	int max_clock, clock, i;
 
@@ -2613,15 +2670,14 @@ sdhci_cam_get_possible_host_clock(struct sdhci_slot *s
 	return clock;
 }
 
-int
+static int
 sdhci_cam_settran_settings(struct sdhci_slot *slot, union ccb *ccb)
 {
 	struct mmc_ios *ios;
-	struct mmc_ios *new_ios;
-	struct ccb_trans_settings_mmc *cts;
+	const struct mmc_ios *new_ios;
+	const struct ccb_trans_settings_mmc *cts;
 
 	ios = &slot->host.ios;
-
 	cts = &ccb->cts.proto_specific.mmc;
 	new_ios = &cts->ios;
 
@@ -2659,7 +2715,7 @@ sdhci_cam_settran_settings(struct sdhci_slot *slot, un
 	return (sdhci_cam_update_ios(slot));
 }
 
-int
+static int
 sdhci_cam_update_ios(struct sdhci_slot *slot)
 {
 	struct mmc_ios *ios = &slot->host.ios;
@@ -2701,10 +2757,10 @@ sdhci_cam_update_ios(struct sdhci_slot *slot)
 	return (0);
 }
 
-int
+static int
 sdhci_cam_request(struct sdhci_slot *slot, union ccb *ccb)
 {
-	struct ccb_mmcio *mmcio;
+	const struct ccb_mmcio *mmcio;
 
 	mmcio = &ccb->mmcio;
 
@@ -2739,4 +2795,4 @@ sdhci_cam_request(struct sdhci_slot *slot, union ccb *
 }
 #endif /* MMCCAM */
 
-MODULE_VERSION(sdhci, 1);
+MODULE_VERSION(sdhci, SDHCI_VERSION);

Modified: stable/12/sys/dev/sdhci/sdhci.h
==============================================================================
--- stable/12/sys/dev/sdhci/sdhci.h	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/sdhci.h	Sun Jan 27 19:04:02 2019	(r343503)
@@ -32,8 +32,8 @@
 
 #include "opt_mmccam.h"
 
-#define	DMA_BLOCK_SIZE	4096
-#define	DMA_BOUNDARY	0	/* DMA reload every 4K */
+/* Macro for sizing the SDMA bounce buffer on the SDMA buffer boundary. */
+#define	SDHCI_SDMA_BNDRY_TO_BBUFSZ(bndry)	(4096 * (1 << bndry))
 
 /* Controller doesn't honor resets unless we touch the clock register */
 #define	SDHCI_QUIRK_CLOCK_BEFORE_RESET			(1 << 0)
@@ -95,6 +95,8 @@
 #define	SDHCI_QUIRK_BROKEN_AUTO_STOP			(1 << 28)
 /* Controller supports eMMC HS400 mode if SDHCI_CAN_SDR104 is set. */
 #define	SDHCI_QUIRK_MMC_HS400_IF_CAN_SDR104		(1 << 29)
+/* SDMA boundary in SDHCI_BLOCK_SIZE broken - use front-end supplied value. */
+#define	SDHCI_QUIRK_BROKEN_SDMA_BOUNDARY		(1 << 30)
 
 /*
  * Controller registers
@@ -102,6 +104,14 @@
 #define	SDHCI_DMA_ADDRESS	0x00
 
 #define	SDHCI_BLOCK_SIZE	0x04
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_4K	0x00
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_8K	0x01
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_16K	0x02
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_32K	0x03
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_64K	0x04
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_128K	0x05
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_256K	0x06
+#define	 SDHCI_BLKSZ_SDMA_BNDRY_512K	0x07
 #define	 SDHCI_MAKE_BLKSZ(dma, blksz) (((dma & 0x7) << 12) | (blksz & 0xFFF))
 
 #define	SDHCI_BLOCK_COUNT	0x06
@@ -362,6 +372,8 @@ struct sdhci_slot {
 	bus_dmamap_t	dmamap;
 	u_char		*dmamem;
 	bus_addr_t	paddr;		/* DMA buffer address */
+	uint32_t	sdma_bbufsz;	/* SDMA bounce buffer size */
+	uint8_t		sdma_boundary;	/* SDMA boundary */
 	struct task	card_task;	/* Card presence check task */
 	struct timeout_task
 			card_delayed_task;/* Card insert delayed task */
@@ -433,5 +445,10 @@ uint32_t sdhci_generic_min_freq(device_t brdev, struct
 bool sdhci_generic_get_card_present(device_t brdev, struct sdhci_slot *slot);
 void sdhci_generic_set_uhs_timing(device_t brdev, struct sdhci_slot *slot);
 void sdhci_handle_card_present(struct sdhci_slot *slot, bool is_present);
+
+#define	SDHCI_VERSION	2
+
+#define	SDHCI_DEPEND(name)						\
+    MODULE_DEPEND(name, sdhci, SDHCI_VERSION, SDHCI_VERSION, SDHCI_VERSION);
 
 #endif	/* __SDHCI_H__ */

Modified: stable/12/sys/dev/sdhci/sdhci_acpi.c
==============================================================================
--- stable/12/sys/dev/sdhci/sdhci_acpi.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/sdhci_acpi.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -456,7 +456,7 @@ static devclass_t sdhci_acpi_devclass;
 
 DRIVER_MODULE(sdhci_acpi, acpi, sdhci_acpi_driver, sdhci_acpi_devclass, NULL,
     NULL);
-MODULE_DEPEND(sdhci_acpi, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_acpi);
 
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_acpi);

Modified: stable/12/sys/dev/sdhci/sdhci_fdt.c
==============================================================================
--- stable/12/sys/dev/sdhci/sdhci_fdt.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/sdhci_fdt.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -80,6 +80,7 @@ struct sdhci_fdt_softc {
 	u_int		quirks;		/* Chip specific quirks */
 	u_int		caps;		/* If we override SDHCI_CAPABILITIES */
 	uint32_t	max_clk;	/* Max possible freq */
+	uint8_t		sdma_boundary;	/* If we override the SDMA boundary */
 	struct resource *irq_res;	/* IRQ resource */
 	void		*intrhand;	/* Interrupt handle */
 
@@ -206,7 +207,9 @@ sdhci_fdt_probe(device_t dev)
 		device_set_desc(dev, "generic fdt SDHCI controller");
 		break;
 	case SDHCI_FDT_QUALCOMM:
-		sc->quirks = SDHCI_QUIRK_ALL_SLOTS_NON_REMOVABLE;
+		sc->quirks = SDHCI_QUIRK_ALL_SLOTS_NON_REMOVABLE |
+		    SDHCI_QUIRK_BROKEN_SDMA_BOUNDARY;
+		sc->sdma_boundary = SDHCI_BLKSZ_SDMA_BNDRY_4K;
 		device_set_desc(dev, "Qualcomm FDT SDHCI controller");
 		break;
 	case SDHCI_FDT_XLNX_ZY7:
@@ -271,6 +274,7 @@ sdhci_fdt_attach(device_t dev)
 		slot->quirks = sc->quirks;
 		slot->caps = sc->caps;
 		slot->max_clk = sc->max_clk;
+		slot->sdma_boundary = sc->sdma_boundary;
 
 		if (sdhci_init_slot(dev, slot, i) != 0)
 			continue;
@@ -353,7 +357,7 @@ static devclass_t sdhci_fdt_devclass;
 
 DRIVER_MODULE(sdhci_fdt, simplebus, sdhci_fdt_driver, sdhci_fdt_devclass,
     NULL, NULL);
-MODULE_DEPEND(sdhci_fdt, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_fdt);
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_fdt);
 #endif

Modified: stable/12/sys/dev/sdhci/sdhci_pci.c
==============================================================================
--- stable/12/sys/dev/sdhci/sdhci_pci.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/sdhci_pci.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -526,7 +526,7 @@ static devclass_t sdhci_pci_devclass;
 
 DRIVER_MODULE(sdhci_pci, pci, sdhci_pci_driver, sdhci_pci_devclass, NULL,
     NULL);
-MODULE_DEPEND(sdhci_pci, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_pci);
 
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_pci);

Modified: stable/12/sys/dev/sdhci/sdhci_xenon.c
==============================================================================
--- stable/12/sys/dev/sdhci/sdhci_xenon.c	Sun Jan 27 18:53:36 2019	(r343502)
+++ stable/12/sys/dev/sdhci/sdhci_xenon.c	Sun Jan 27 19:04:02 2019	(r343503)
@@ -545,7 +545,7 @@ static devclass_t sdhci_xenon_devclass;
 DRIVER_MODULE(sdhci_xenon, simplebus, sdhci_xenon_driver, sdhci_xenon_devclass,
     NULL, NULL);
 
-MODULE_DEPEND(sdhci_xenon, sdhci, 1, 1, 1);
+SDHCI_DEPEND(sdhci_xenon);
 #ifndef MMCCAM
 MMC_DECLARE_BRIDGE(sdhci_xenon);
 #endif



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201901271904.x0RJ4264085105>