lp/drivers/usb: Work around QEMU XHCI register issue

The QEMU XHCI controller does not support byte/word reads from the
capability register and it expects dword reads only.

In order to make this work move the access of the capability
register fields to use macros instead of a packed struct bitfield.

This issue was filed upstream:
https://bugs.launchpad.net/qemu/+bug/1693050

The original fix attempt in 2012 was not effective:
https://github.com/qemu/qemu/commit/6ee021d41078844df60a3a466e3829a3e82776f3

With this change the controller is detected properly by the libpayload
USB drivers.

Change-Id: I048ed14921a4c9c0620c10b315b42476b6e5c512
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/39838
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>

3 files changed, 81 insertions, 78 deletions

diff --git a/payloads/libpayload/drivers/usb/xhci.c b/payloads/libpayload/drivers/usb/xhci.c
index 0a69c5137b..2f61f8658a 100644
--- a/payloads/libpayload/drivers/usb/xhci.c
+++ b/payloads/libpayload/drivers/usb/xhci.c
@@ -185,26 +185,27 @@ xhci_init (unsigned long physical_bar)
 		goto _free_xhci;
 	}
 
-	xhci->capreg	= phys_to_virt(physical_bar);
-	xhci->opreg	= ((void *)xhci->capreg) + xhci->capreg->caplength;
-	xhci->hcrreg	= ((void *)xhci->capreg) + xhci->capreg->rtsoff;
-	xhci->dbreg	= ((void *)xhci->capreg) + xhci->capreg->dboff;
+	memcpy(&xhci->capreg, phys_to_virt(physical_bar), sizeof(xhci->capreg));
+	xhci->opreg = phys_to_virt(physical_bar) + CAP_GET(CAPLEN, xhci->capreg);
+	xhci->hcrreg = phys_to_virt(physical_bar) + xhci->capreg.rtsoff;
+	xhci->dbreg = phys_to_virt(physical_bar) + xhci->capreg.dboff;
+
 	xhci_debug("regbase: 0x%"PRIx32"\n", physical_bar);
-	xhci_debug("caplen:  0x%"PRIx32"\n", xhci->capreg->caplength);
-	xhci_debug("rtsoff:  0x%"PRIx32"\n", xhci->capreg->rtsoff);
-	xhci_debug("dboff:   0x%"PRIx32"\n", xhci->capreg->dboff);
+	xhci_debug("caplen:  0x%"PRIx32"\n", CAP_GET(CAPLEN, xhci->capreg));
+	xhci_debug("rtsoff:  0x%"PRIx32"\n", xhci->capreg.rtsoff);
+	xhci_debug("dboff:   0x%"PRIx32"\n", xhci->capreg.dboff);
 
 	xhci_debug("hciversion: %"PRIx8".%"PRIx8"\n",
-		   xhci->capreg->hciver_hi, xhci->capreg->hciver_lo);
-	if ((xhci->capreg->hciversion < 0x96) ||
-			(xhci->capreg->hciversion > 0x110)) {
+		   CAP_GET(CAPVER_HI, xhci->capreg), CAP_GET(CAPVER_LO, xhci->capreg));
+	if ((CAP_GET(CAPVER, xhci->capreg) < 0x96) ||
+	    (CAP_GET(CAPVER, xhci->capreg) > 0x110)) {
 		xhci_debug("Unsupported xHCI version\n");
 		goto _free_xhci;
 	}
 
 	xhci_debug("context size: %dB\n", CTXSIZE(xhci));
-	xhci_debug("maxslots: 0x%02lx\n", xhci->capreg->MaxSlots);
-	xhci_debug("maxports: 0x%02lx\n", xhci->capreg->MaxPorts);
+	xhci_debug("maxslots: 0x%02lx\n", CAP_GET(MAXSLOTS, xhci->capreg));
+	xhci_debug("maxports: 0x%02lx\n", CAP_GET(MAXPORTS, xhci->capreg));
 	const unsigned pagesize = xhci->opreg->pagesize << 12;
 	xhci_debug("pagesize: 0x%04x\n", pagesize);
 
@@ -213,7 +214,8 @@ xhci_init (unsigned long physical_bar)
 	 * structures at first and can still chicken out easily if we run out
 	 * of memory.
 	 */
-	xhci->max_slots_en = xhci->capreg->MaxSlots & CONFIG_LP_MASK_MaxSlotsEn;
+	xhci->max_slots_en = CAP_GET(MAXSLOTS, xhci->capreg) &
+		CONFIG_LP_MASK_MaxSlotsEn;
 	xhci->dcbaa = xhci_align(64, (xhci->max_slots_en + 1) * sizeof(u64));
 	xhci->dev = malloc((xhci->max_slots_en + 1) * sizeof(*xhci->dev));
 	if (!xhci->dcbaa || !xhci->dev) {
@@ -227,8 +229,9 @@ xhci_init (unsigned long physical_bar)
 	 * Let dcbaa[0] point to another array of pointers, sp_ptrs.
 	 * The pointers therein point to scratchpad buffers (pages).
 	 */
-	const size_t max_sp_bufs = xhci->capreg->Max_Scratchpad_Bufs_Hi << 5 |
-				   xhci->capreg->Max_Scratchpad_Bufs_Lo;
+	const size_t max_sp_bufs =
+		CAP_GET(MAX_SCRATCH_BUFS_HI, xhci->capreg) << 5 |
+		CAP_GET(MAX_SCRATCH_BUFS_LO, xhci->capreg);
 	xhci_debug("max scratchpad bufs: 0x%zx\n", max_sp_bufs);
 	if (max_sp_bufs) {
 		const size_t sp_ptrs_size = max_sp_bufs * sizeof(u64);
@@ -376,7 +379,8 @@ xhci_reinit (hci_t *controller)
 	xhci_debug("event ring @%p (0x%08x)\n",
 		   xhci->er.ring, virt_to_phys(xhci->er.ring));
 	xhci_debug("ERST Max: 0x%lx ->  0x%lx entries\n",
-		   xhci->capreg->ERST_Max, 1 << xhci->capreg->ERST_Max);
+		   CAP_GET(ERST_MAX, xhci->capreg),
+		   1 << CAP_GET(ERST_MAX, xhci->capreg));
 	memset((void*)xhci->ev_ring_table, 0x00, sizeof(erst_entry_t));
 	xhci->ev_ring_table[0].seg_base_lo = virt_to_phys(xhci->er.ring);
 	xhci->ev_ring_table[0].seg_base_hi = 0;
@@ -432,8 +436,9 @@ xhci_shutdown(hci_t *const controller)
 #endif
 
 	if (xhci->sp_ptrs) {
-		size_t max_sp_bufs = xhci->capreg->Max_Scratchpad_Bufs_Hi << 5 |
-				     xhci->capreg->Max_Scratchpad_Bufs_Lo;
+		const size_t max_sp_bufs =
+			CAP_GET(MAX_SCRATCH_BUFS_HI, xhci->capreg) << 5 |
+			CAP_GET(MAX_SCRATCH_BUFS_LO, xhci->capreg);
 		for (i = 0; i < max_sp_bufs; ++i) {
 			if (xhci->sp_ptrs[i])
 				free(phys_to_virt(xhci->sp_ptrs[i]));
diff --git a/payloads/libpayload/drivers/usb/xhci_private.h b/payloads/libpayload/drivers/usb/xhci_private.h
index ab1dfa98e1..65c3fdd6cc 100644
--- a/payloads/libpayload/drivers/usb/xhci_private.h
+++ b/payloads/libpayload/drivers/usb/xhci_private.h
@@ -274,7 +274,6 @@ typedef volatile struct epctx {
 } epctx_t;
 
 #define NUM_EPS 32
-#define CTXSIZE(xhci) ((xhci)->capreg->csz ? 64 : 32)
 
 typedef union devctx {
 	/* set of pointers, so we can dynamically adjust Slot/EP context size */
@@ -321,66 +320,65 @@ typedef struct erst_entry {
 	u32 rsvd;
 } erst_entry_t;
 
+#define CAP_CAPLEN_FIELD		hciparams
+#define CAP_CAPLEN_START		0
+#define CAP_CAPLEN_LEN			8
+#define CAP_CAPVER_FIELD		hciparams
+#define CAP_CAPVER_START		16
+#define CAP_CAPVER_LEN			16
+#define CAP_CAPVER_HI_FIELD		hciparams
+#define CAP_CAPVER_HI_START		24
+#define CAP_CAPVER_HI_LEN		8
+#define CAP_CAPVER_LO_FIELD		hciparams
+#define CAP_CAPVER_LO_START		16
+#define CAP_CAPVER_LO_LEN		8
+#define CAP_MAXSLOTS_FIELD		hcsparams1
+#define CAP_MAXSLOTS_START		0
+#define CAP_MAXSLOTS_LEN		7
+#define CAP_MAXINTRS_FIELD		hcsparams1
+#define CAP_MAXINTRS_START		7
+#define CAP_MAXINTRS_LEN		11
+#define CAP_MAXPORTS_FIELD		hcsparams1
+#define CAP_MAXPORTS_START		24
+#define CAP_MAXPORTS_LEN		8
+#define CAP_IST_FIELD			hcsparams2
+#define CAP_IST_START			0
+#define CAP_IST_LEN			4
+#define CAP_ERST_MAX_FIELD		hcsparams2
+#define CAP_ERST_MAX_START		4
+#define CAP_ERST_MAX_LEN		4
+#define CAP_MAX_SCRATCH_BUFS_HI_FIELD	hcsparams2
+#define CAP_MAX_SCRATCH_BUFS_HI_START	21
+#define CAP_MAX_SCRATCH_BUFS_HI_LEN	5
+#define CAP_MAX_SCRATCH_BUFS_LO_FIELD	hcsparams2
+#define CAP_MAX_SCRATCH_BUFS_LO_START	27
+#define CAP_MAX_SCRATCH_BUFS_LO_LEN	5
+#define CAP_U1_LATENCY_FIELD		hcsparams3
+#define CAP_U1_LATENCY_START		0
+#define CAP_U1_LATENCY_LEN		8
+#define CAP_U2_LATENCY_FIELD		hcsparams3
+#define CAP_U2_LATENCY_START		16
+#define CAP_U2_LATENCY_LEN		16
+#define CAP_CSZ_FIELD			hccparams
+#define CAP_CSZ_START			2
+#define CAP_CSZ_LEN			1
+
+#define CAP_MASK(tok)		MASK(CAP_##tok##_START, CAP_##tok##_LEN)
+#define CAP_GET(tok, cap)	(((cap).CAP_##tok##_FIELD & CAP_MASK(tok)) \
+				 >> CAP_##tok##_START)
+
+#define CTXSIZE(xhci) (CAP_GET(CSZ, (xhci)->capreg) ? 64 : 32)
+
 typedef struct xhci {
-	/* capreg is read-only, so no need for volatile,
-	   and thus 32bit accesses can be assumed. */
 	struct capreg {
-		u8 caplength; /* 0x00 */
-		u8 res1; /* 0x01 */
-		union { /* 0x02 */
-			u16 hciversion;
-			struct {
-				u8 hciver_lo;
-				u8 hciver_hi;
-			} __packed;
-		} __packed;
-		union { /* 0x04 */
-			u32 hcsparams1;
-			struct {
-				unsigned long MaxSlots:7;
-				unsigned long MaxIntrs:11;
-				unsigned long:6;
-				unsigned long MaxPorts:8;
-			} __packed;
-		} __packed;
-		union { /* 0x08 */
-			u32 hcsparams2;
-			struct {
-				unsigned long IST:4;
-				unsigned long ERST_Max:4;
-				unsigned long:13;
-				unsigned long Max_Scratchpad_Bufs_Hi:5;
-				unsigned long SPR:1;
-				unsigned long Max_Scratchpad_Bufs_Lo:5;
-			} __packed;
-		} __packed;
-		union { /* 0x0C */
-			u32 hcsparams3;
-			struct {
-				unsigned long u1latency:8;
-				unsigned long:8;
-				unsigned long u2latency:16;
-			} __packed;
-		} __packed;
-		union { /* 0x10 */
-			u32 hccparams;
-			struct {
-				unsigned long ac64:1;
-				unsigned long bnc:1;
-				unsigned long csz:1;
-				unsigned long ppc:1;
-				unsigned long pind:1;
-				unsigned long lhrc:1;
-				unsigned long ltc:1;
-				unsigned long nss:1;
-				unsigned long:4;
-				unsigned long MaxPSASize:4;
-				unsigned long xECP:16;
-			} __packed;
-		} __packed;
-		u32 dboff; /* 0x14 */
-		u32 rtsoff; /* 0x18 */
-	} __packed *capreg;
+		u32 hciparams;
+		u32 hcsparams1;
+		u32 hcsparams2;
+		u32 hcsparams3;
+		u32 hccparams;
+		u32 dboff;
+		u32 rtsoff;
+	} __packed capreg;
 
 	/* opreg is R/W is most places, so volatile access is necessary.
 	   volatile means that the compiler seeks byte writes if possible,
diff --git a/payloads/libpayload/drivers/usb/xhci_rh.c b/payloads/libpayload/drivers/usb/xhci_rh.c
index 453fa5b409..865b9ac18b 100644
--- a/payloads/libpayload/drivers/usb/xhci_rh.c
+++ b/payloads/libpayload/drivers/usb/xhci_rh.c
@@ -160,7 +160,7 @@ xhci_rh_init (usbdev_t *dev)
 	dev->port = -1;
 
 	const int num_ports = /* TODO: maybe we need to read extended caps */
-		(XHCI_INST(dev->controller)->capreg->hcsparams1 >> 24) & 0xff;
+		CAP_GET(MAXPORTS, XHCI_INST(dev->controller)->capreg);
 	generic_hub_init(dev, num_ports, &xhci_rh_ops);
 
 	usb_debug("xHCI: root hub init done\n");