Merge branch 'CR_9958_6.6_usb_perf_minda' into 'jh7110-6.6.y-devel'

CR_9958 usb:xhci:To improve performance,usb using lowmem for bulk xfer.

See merge request sdk/linux!1037

arch/riscv/boot/dts/starfive/jh7110-evb.dts

@@ -31,5 +31,6 @@
 };
 
 &usb0 {
+	xhci-lowmem-pool;
 	status = "okay";
 };

drivers/usb/core/hcd.c

@@ -1439,7 +1439,9 @@ 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 ! */

drivers/usb/host/xhci-mem.c

@@ -14,6 +14,7 @@
 #include <linux/slab.h>
 #include <linux/dmapool.h>
 #include <linux/dma-mapping.h>
+#include <linux/genalloc.h>
 
 #include "xhci.h"
 #include "xhci-trace.h"
@@ -1840,6 +1841,7 @@ xhci_free_interrupter(struct xhci_hcd *xhci, struct xhci_interrupter *ir)
 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);
@@ -1885,6 +1887,14 @@ 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_noncoherent(dev, pool->size, (void *)pool->cached_base,
+				     pool->dma_addr, DMA_BIDIRECTIONAL);
+		gen_pool_destroy(pool->pool);
+		pool->pool = NULL;
+	}
+
 	if (xhci->dcbaa)
 		dma_free_coherent(dev, sizeof(*xhci->dcbaa),
 				xhci->dcbaa, xhci->dcbaa->dma);
@@ -2295,6 +2305,55 @@ xhci_add_interrupter(struct xhci_hcd *xhci, struct xhci_interrupter *ir,
 	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 (!pool->pool)
+			return -ENOMEM;
+	}
+
+	buffer = dma_alloc_noncoherent(hcd->self.sysdev, size, &dma_addr,
+		DMA_BIDIRECTIONAL, GFP_ATOMIC);
+
+	if (!buffer) {
+		err = -ENOMEM;
+		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_noncoherent(hcd->self.sysdev, size, buffer, dma_addr, DMA_BIDIRECTIONAL);
+		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;
@@ -2302,7 +2361,7 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	unsigned int	val, val2;
 	u64		val_64;
 	u32		page_size, temp;
-	int		i;
+	int		i, ret;
 
 	INIT_LIST_HEAD(&xhci->cmd_list);
 
@@ -2431,6 +2490,14 @@ 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) {
+		ret = xhci_setup_local_lowmem(xhci, xhci->lowmem_pool.size);
+		if (ret) {
+			xhci->quirks &= ~XHCI_LOCAL_BUFFER;
+			xhci_warn(xhci, "WARN: Can't alloc lowmem pool\n");
+		}
+	}
+
 	/*
 	 * XXX: Might need to set the Interrupter Moderation Register to
 	 * something other than the default (~1ms minimum between interrupts).

drivers/usb/host/xhci-plat.c

@@ -253,6 +253,15 @@ int xhci_plat_probe(struct platform_device *pdev, struct device *sysdev, const s
 		if (device_property_read_bool(tmpdev, "xhci-sg-trb-cache-size-quirk"))
 			xhci->quirks |= XHCI_SG_TRB_CACHE_SIZE_QUIRK;
 
+		if (device_property_read_bool(tmpdev, "xhci-lowmem-pool")) {
+			xhci->quirks |= XHCI_LOCAL_BUFFER;
+			ret = device_property_read_u32(tmpdev, "lowmem-pool-size",
+				&xhci->lowmem_pool.size);
+			if (ret || xhci->lowmem_pool.size >= 4)
+				xhci->lowmem_pool.size = 4 << 20;
+			else
+				xhci->lowmem_pool.size <<= 20;
+		}
 		device_property_read_u32(tmpdev, "imod-interval-ns",
 					 &xhci->imod_interval);
 	}

drivers/usb/host/xhci-ring.c

@@ -3612,7 +3612,8 @@ 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);

drivers/usb/host/xhci.c

@@ -18,6 +18,8 @@
 #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"
@@ -1273,6 +1275,55 @@ 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
@@ -1293,9 +1344,11 @@ 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;
@@ -1308,8 +1361,10 @@ 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);
+	}
 }
 
 /**

drivers/usb/host/xhci.h

@@ -1757,6 +1757,13 @@ 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;
@@ -1908,6 +1915,8 @@ struct xhci_hcd {
 #define XHCI_ZHAOXIN_TRB_FETCH	BIT_ULL(45)
 #define XHCI_ZHAOXIN_HOST	BIT_ULL(46)
 
+#define XHCI_LOCAL_BUFFER	BIT_ULL(63)
+
 	unsigned int		num_active_eps;
 	unsigned int		limit_active_eps;
 	struct xhci_port	*hw_ports;
@@ -1937,6 +1946,8 @@ 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));
 };