usb:xhci:To improve performance,usb using lowmem for bulk xfer.

Generate an usb low memory pool for usb 3.0 host
read/write transfer, default size is 8M.

Signed-off-by: minda.chen <minda.chen@starfivetech.com>

7 files changed, 145 insertions, 3 deletions

diff --git a/arch/riscv/boot/dts/starfive/jh7110-evb.dts b/arch/riscv/boot/dts/starfive/jh7110-evb.dts
index d110913947f7..d6345d01964a 100644
--- a/arch/riscv/boot/dts/starfive/jh7110-evb.dts
+++ b/arch/riscv/boot/dts/starfive/jh7110-evb.dts
@@ -31,6 +31,7 @@
 };
 
 &usbdrd30 {
+	xhci-lowmem-pool;
 	status = "okay";
 };
 
diff --git a/drivers/usb/core/hcd.c b/drivers/usb/core/hcd.c
index 7ee6e4cc0d89..1b6c8bf11aac 100644
--- a/drivers/usb/core/hcd.c
+++ b/drivers/usb/core/hcd.c
@@ -1452,7 +1452,9 @@ int usb_hcd_map_urb_for_dma(struct usb_hcd *hcd, struct urb *urb,
 			if (ret == 0)
 				urb->transfer_flags |= URB_MAP_LOCAL;
 		} else if (hcd_uses_dma(hcd)) {
-			if (urb->num_sgs) {
+			if (urb->transfer_flags & URB_MAP_LOCAL)
+				return ret;
+			else if (urb->num_sgs) {
 				int n;
 
 				/* We don't support sg for isoc transfers ! */
diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
index 0e312066c5c6..d8ca7e694ca3 100644
--- a/drivers/usb/host/xhci-mem.c
+++ b/drivers/usb/host/xhci-mem.c
@@ -13,6 +13,7 @@
 #include <linux/slab.h>
 #include <linux/dmapool.h>
 #include <linux/dma-mapping.h>
+#include <linux/genalloc.h>
 
 #include "xhci.h"
 #include "xhci-trace.h"
@@ -1835,6 +1836,7 @@ void xhci_free_erst(struct xhci_hcd *xhci, struct xhci_erst *erst)
 void xhci_mem_cleanup(struct xhci_hcd *xhci)
 {
 	struct device	*dev = xhci_to_hcd(xhci)->self.sysdev;
+	struct xhci_lowmem_pool *pool;
 	int i, j, num_ports;
 
 	cancel_delayed_work_sync(&xhci->cmd_timer);
@@ -1886,6 +1888,13 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 	xhci_dbg_trace(xhci, trace_xhci_dbg_init,
 			"Freed medium stream array pool");
 
+	if (xhci->lowmem_pool.pool) {
+		pool = &xhci->lowmem_pool;
+		dma_free_coherent(dev, pool->size, (void *)pool->cached_base, pool->dma_addr);
+		gen_pool_destroy(pool->pool);
+		pool->pool = NULL;
+	}
+
 	if (xhci->dcbaa)
 		dma_free_coherent(dev, sizeof(*xhci->dcbaa),
 				xhci->dcbaa, xhci->dcbaa->dma);
@@ -2377,6 +2386,55 @@ static int xhci_setup_port_arrays(struct xhci_hcd *xhci, gfp_t flags)
 	return 0;
 }
 
+int xhci_setup_local_lowmem(struct xhci_hcd *xhci, size_t size)
+{
+	int err;
+	void *buffer;
+	dma_addr_t dma_addr;
+	struct usb_hcd *hcd = xhci_to_hcd(xhci);
+	struct xhci_lowmem_pool *pool = &xhci->lowmem_pool;
+
+	if (!pool->pool) {
+		/* minimal alloc one page */
+		pool->pool = gen_pool_create(PAGE_SHIFT, dev_to_node(hcd->self.sysdev));
+		if (IS_ERR(pool->pool))
+			return PTR_ERR(pool->pool);
+	}
+
+	buffer = dma_alloc_coherent(hcd->self.sysdev, size, &dma_addr,
+			GFP_KERNEL | GFP_DMA32);
+
+	if (IS_ERR(buffer)) {
+		err = PTR_ERR(buffer);
+		goto destroy_pool;
+	}
+
+	/*
+	 * Here we pass a dma_addr_t but the arg type is a phys_addr_t.
+	 * It's not backed by system memory and thus there's no kernel mapping
+	 * for it.
+	 */
+	err = gen_pool_add_virt(pool->pool, (unsigned long)buffer,
+				dma_addr, size, dev_to_node(hcd->self.sysdev));
+	if (err < 0) {
+		dev_err(hcd->self.sysdev, "gen_pool_add_virt failed with %d\n",
+			err);
+		dma_free_coherent(hcd->self.sysdev, size, buffer, dma_addr);
+		goto destroy_pool;
+	}
+
+	pool->cached_base = (u64)buffer;
+	pool->dma_addr = dma_addr;
+
+	return 0;
+
+destroy_pool:
+	gen_pool_destroy(pool->pool);
+	pool->pool = NULL;
+	return err;
+}
+EXPORT_SYMBOL_GPL(xhci_setup_local_lowmem);
+
 int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 {
 	dma_addr_t	dma;
@@ -2550,6 +2608,12 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 
 	xhci->isoc_bei_interval = AVOID_BEI_INTERVAL_MAX;
 
+	if (xhci->quirks & XHCI_LOCAL_BUFFER) {
+		if (xhci_setup_local_lowmem(xhci,
+			xhci->lowmem_pool.size))
+			goto fail;
+	}
+
 	/*
 	 * XXX: Might need to set the Interrupter Moderation Register to
 	 * something other than the default (~1ms minimum between interrupts).
diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c
index c1edcc9b13ce..855f166dd1ae 100644
--- a/drivers/usb/host/xhci-plat.c
+++ b/drivers/usb/host/xhci-plat.c
@@ -323,6 +323,14 @@ static int xhci_plat_probe(struct platform_device *pdev)
 		if (device_property_read_bool(tmpdev, "quirk-broken-port-ped"))
 			xhci->quirks |= XHCI_BROKEN_PORT_PED;
 
+		if (device_property_read_bool(tmpdev, "xhci-lowmem-pool")) {
+			xhci->quirks |= XHCI_LOCAL_BUFFER;
+			if (device_property_read_u32(tmpdev, "lowmem-pool-size",
+				&xhci->lowmem_pool.size)) {
+				xhci->lowmem_pool.size = 8 << 20;
+			} else
+				xhci->lowmem_pool.size <<= 20;
+		}
 		device_property_read_u32(tmpdev, "imod-interval-ns",
 					 &xhci->imod_interval);
 	}
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 311597bba80e..ed97839c105a 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -3592,7 +3592,8 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	full_len = urb->transfer_buffer_length;
 	/* If we have scatter/gather list, we use it. */
-	if (urb->num_sgs && !(urb->transfer_flags & URB_DMA_MAP_SINGLE)) {
+	if (urb->num_sgs && !(urb->transfer_flags & URB_DMA_MAP_SINGLE)
+			&& !(urb->transfer_flags & URB_MAP_LOCAL)) {
 		num_sgs = urb->num_mapped_sgs;
 		sg = urb->sg;
 		addr = (u64) sg_dma_address(sg);
diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 541fe4dcc43a..33eb3d52e376 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -17,6 +17,8 @@
 #include <linux/slab.h>
 #include <linux/dmi.h>
 #include <linux/dma-mapping.h>
+#include <linux/dma-map-ops.h>
+#include <linux/genalloc.h>
 
 #include "xhci.h"
 #include "xhci-trace.h"
@@ -1380,6 +1382,55 @@ static void xhci_unmap_temp_buf(struct usb_hcd *hcd, struct urb *urb)
 	urb->transfer_buffer = NULL;
 }
 
+static void xhci_map_urb_local(struct usb_hcd *hcd, struct urb *urb,
+				gfp_t mem_flags)
+{
+	void *buffer;
+	dma_addr_t dma_handle;
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct xhci_lowmem_pool *lowmem_pool = &xhci->lowmem_pool;
+
+	if (lowmem_pool->pool
+		&& (usb_endpoint_type(&urb->ep->desc) == USB_ENDPOINT_XFER_BULK)
+		&& (urb->transfer_buffer_length > PAGE_SIZE)
+		&& urb->num_sgs && urb->sg && (sg_phys(urb->sg) > 0xffffffff)) {
+		buffer = gen_pool_dma_alloc(lowmem_pool->pool,
+			urb->transfer_buffer_length, &dma_handle);
+		if (buffer) {
+			urb->transfer_dma = dma_handle;
+			urb->transfer_buffer = buffer;
+			urb->transfer_flags |= URB_MAP_LOCAL;
+			if (usb_urb_dir_out(urb))
+				sg_copy_to_buffer(urb->sg, urb->num_sgs,
+					(void *)buffer,
+					urb->transfer_buffer_length);
+		}
+	}
+
+}
+
+static void xhci_unmap_urb_local(struct usb_hcd *hcd, struct urb *urb)
+{
+	dma_addr_t dma_handle;
+	u64 cached_buffer;
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct xhci_lowmem_pool *lowmem_pool = &xhci->lowmem_pool;
+
+	if (urb->transfer_flags & URB_MAP_LOCAL) {
+		dma_handle = urb->transfer_dma;
+		cached_buffer = lowmem_pool->cached_base +
+			((u32)urb->transfer_dma & (lowmem_pool->size - 1));
+		if (usb_urb_dir_in(urb))
+			sg_copy_from_buffer(urb->sg, urb->num_sgs,
+				(void *)cached_buffer, urb->transfer_buffer_length);
+		gen_pool_free(lowmem_pool->pool, (unsigned long)urb->transfer_buffer,
+				urb->transfer_buffer_length);
+		urb->transfer_flags &= ~URB_MAP_LOCAL;
+		urb->transfer_buffer = NULL;
+	}
+}
+
+
 /*
  * Bypass the DMA mapping if URB is suitable for Immediate Transfer (IDT),
  * we'll copy the actual data into the TRB address register. This is limited to
@@ -1400,9 +1451,11 @@ static int xhci_map_urb_for_dma(struct usb_hcd *hcd, struct urb *urb,
 		if (xhci_urb_temp_buffer_required(hcd, urb))
 			return xhci_map_temp_buffer(hcd, urb);
 	}
+	xhci_map_urb_local(hcd, urb, mem_flags);
 	return usb_hcd_map_urb_for_dma(hcd, urb, mem_flags);
 }
 
+
 static void xhci_unmap_urb_for_dma(struct usb_hcd *hcd, struct urb *urb)
 {
 	struct xhci_hcd *xhci;
@@ -1415,8 +1468,10 @@ static void xhci_unmap_urb_for_dma(struct usb_hcd *hcd, struct urb *urb)
 
 	if ((xhci->quirks & XHCI_SG_TRB_CACHE_SIZE_QUIRK) && unmap_temp_buf)
 		xhci_unmap_temp_buf(hcd, urb);
-	else
+	else {
+		xhci_unmap_urb_local(hcd, urb);
 		usb_hcd_unmap_urb_for_dma(hcd, urb);
+	}
 }
 
 /**
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index 5a75fe563123..827bc2c32ff9 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -1747,6 +1747,13 @@ struct xhci_hub {
 	u8			min_rev;
 };
 
+struct xhci_lowmem_pool {
+	struct gen_pool *pool;
+	u64		cached_base;
+	dma_addr_t	dma_addr;
+	unsigned int	size;
+};
+
 /* There is one xhci_hcd structure per controller */
 struct xhci_hcd {
 	struct usb_hcd *main_hcd;
@@ -1901,6 +1908,8 @@ struct xhci_hcd {
 #define XHCI_BROKEN_D3COLD	BIT_ULL(41)
 #define XHCI_EP_CTX_BROKEN_DCS	BIT_ULL(42)
 
+#define XHCI_LOCAL_BUFFER	BIT_ULL(63)
+
 	unsigned int		num_active_eps;
 	unsigned int		limit_active_eps;
 	struct xhci_port	*hw_ports;
@@ -1928,6 +1937,8 @@ struct xhci_hcd {
 	struct list_head	regset_list;
 
 	void			*dbc;
+	struct xhci_lowmem_pool lowmem_pool;
+
 	/* platform-specific data -- must come last */
 	unsigned long		priv[] __aligned(sizeof(s64));
 };