[riscv] Support T-Head CPUs using non-standard Memory Attribute Extension

Xuantie/T-Head processors such as the C910 (as used in the Sipeed
Lichee Pi 4A) use the high bits of the PTE in a very non-standard way
that is incompatible with the RISC-V specification.

As per the "Memory Attribute Extension (XTheadMae)", bits 62 and 61
represent cacheability and "bufferability" (write-back cacheability)
respectively.  If we do not enable these bits, then the processor gets
incredibly confused at the point that paging is enabled.  The symptom
is that cache lines will occasionally fail to fill, and so reads from
any address may return unrelated data from a previously read cache
line for a different address.

Work around these hardware flaws by detecting T-Head CPUs (via the
"get machine vendor ID" SBI call), then reading the vendor-specific
SXSTATUS register to determine whether or not the vendor-specific
Memory Attribute Extension has been enabled by the M-mode firmware.
If it has, then set bits 61 and 62 in each page table entry that is
used to access normal memory.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/arch/riscv/prefix/libprefix.S

@@ -793,6 +793,33 @@ page_table:
  *      use only a single physical 4kB page table.
  */
 
+/** SBI base extension */
+#define SBI_BASE 0x10
+#define SBI_BASE_MVENDORID 0x04
+
+/** Non-standard T-Head page table entry additional flags
+ *
+ * T-Head processors such as the C910 use the high bits of the PTE in
+ * a very non-standard way that is incompatible with the RISC-V
+ * specification.
+ *
+ * As per the "Memory Attribute Extension (XTheadMae)", bits 62 and 61
+ * represent cacheability and "bufferability" (i.e. write-back
+ * cacheability) respectively.  If we do not enable these bits, then
+ * the processor gets incredibly confused at the point that paging is
+ * enabled.  The symptom is that cache lines will occasionally fail to
+ * fill, and so reads from any address may return unrelated data from
+ * a previously read cache line for a different address.
+ */
+#define THEAD_PTE_MAEE ( 0x60 << ( __riscv_xlen - 8 ) )
+
+/** T-Head vendor ID */
+#define THEAD_MVENDORID 0x5b7
+
+/** T-Head "sxstatus" CSR */
+#define THEAD_CSR_SXSTATUS 0x5c0
+#define THEAD_CSR_SXSTATUS_MAEE	0x00200000	/**< XTheadMae enabled */
+
 	.section ".prefix.enable_paging_64", "ax", @progbits
 enable_paging_64:
 	/* Register usage:
@@ -805,15 +832,39 @@ enable_paging_64:
 	 * a4 - PTE stride
 	 * a5 - size of accessible physical address space
 	 */
-
 	progress " paging:"
-	li	a1, SATP_MODE_SV57
 
 	/* Calculate virtual address offset */
 	LOADN	t0, prefix_link
 	la	t1, _prefix
 	sub	tp, t1, t0
 
+	/* Construct base page table entry for address zero */
+	li	t0, PTE_LEAF
+	STOREN	t0, (a0)
+
+	/* Check for broken T-Head paging extensions */
+	mv	a3, a0
+	li	a7, SBI_BASE
+	li	a6, SBI_BASE_MVENDORID
+	ecall
+	bnez	a0, 1f
+	li	t0, THEAD_MVENDORID
+	bne	a1, t0, 1f
+	progress "thead-"
+	csrr	t0, THEAD_CSR_SXSTATUS
+	li	t1, THEAD_CSR_SXSTATUS_MAEE
+	and	t0, t0, t1
+	beqz	t0, 1f
+	progress "mae-"
+	LOADN	t0, (a3)
+	li	t1, THEAD_PTE_MAEE
+	or	t0, t0, t1
+	STOREN	t0, (a3)
+1:	mv	a0, a3
+
+	/* Find highest supported paging level */
+	li	a1, SATP_MODE_SV57
 enable_paging_64_loop:
 
 	/* Calculate PTE stride for identity map at this paging level
@@ -841,7 +892,7 @@ enable_paging_64_loop:
 	/* Construct PTE[0-255] for identity map */
 	mv	a3, a0
 	li	t0, ( PTE_COUNT / 2 )
-	li	t1, PTE_LEAF
+	LOADN	t1, (a0)
 1:	STOREN	t1, (a3)
 	addi	a3, a3, PTE_SIZE
 	add	t1, t1, a4
@@ -886,13 +937,14 @@ enable_paging_64_loop:
 	/* Construct PTE[x-y] for iPXE virtual address map */
 	la	t0, _prefix
 	srli	t0, t0, PTE_PPN_SHIFT
-	ori	t0, t0, PTE_LEAF
-	la	t1, _ebss
-	srli	t1, t1, PTE_PPN_SHIFT
+	LOADN	t1, (a0)
+	or	t0, t0, t1
+	la	t2, _ebss
+	srli	t2, t2, PTE_PPN_SHIFT
 1:	STOREN	t0, (a3)
 	addi	a3, a3, PTE_SIZE
 	add	t0, t0, a4
-	ble	t0, t1, 1b
+	ble	t0, t2, 1b
 
 	/* Attempt to enable paging, and read back active paging level */
 	slli	t0, a1, SATP_MODE_SHIFT