microblaze: Handle TLB skip size dynamically

This patch fix the problem with rootfs on JFFS2 with early printk
console turned on.

The origin version used TLB63 for temporary early printk mapping.
The code expect that kernel is not able to use all 64 TLB entries
till early printk console is remapped by ioremap. After that
temporary mapping on TLB63 is silently lost.
This expectation give the opportunity to have early console pretty
early.

Microblaze systems with JFFS2 rootfs with early printk console turned on
used more than 64 TLB entries before kernel can remap early console.
Based on that kernel does access to bad area because early printk mapping
is rewritten.

This patch introduces tlb_skip variable which dynamically stores number
of skipped TLB entries from the TLB0. skip_tlb=2 means that TLB0 and TLB1
should be skipped.

MICROBLAZE_TLB_SKIP defines how many TLB is skipped at the kernel start.
They can be used for user purpose.

TLB 63 is used for temporary LMB mapping (MICROBLAZE_LMB_TLB_ID).

Also clean TLBLO when kernel starts.

For specific kernel sizes kernel can use just one TLB. Detect this case
and use the second TLB for general purpose.

Change _tlbia function to flush TLB entries from tlb_skip to TLB_SIZE.

Export tlb_skip size through debugfs.

Signed-off-by: Michal Simek <monstr@monstr.eu>
diff --git a/arch/microblaze/include/asm/mmu.h b/arch/microblaze/include/asm/mmu.h
index 5198de8b..1f9eddd 100644
--- a/arch/microblaze/include/asm/mmu.h
+++ b/arch/microblaze/include/asm/mmu.h
@@ -56,6 +56,12 @@
 
 extern void _tlbie(unsigned long va);	/* invalidate a TLB entry */
 extern void _tlbia(void);		/* invalidate all TLB entries */
+
+/*
+ * tlb_skip size stores actual number skipped TLBs from TLB0 - every directy TLB
+ * mapping has to increase tlb_skip size.
+ */
+extern u32 tlb_skip;
 #   endif /* __ASSEMBLY__ */
 
 /*
@@ -68,7 +74,12 @@
  */
 
 #  define MICROBLAZE_TLB_SIZE 64
-#  define MICROBLAZE_TLB_SKIP 2
+
+/* For cases when you want to skip some TLB entries */
+#  define MICROBLAZE_TLB_SKIP 0
+
+/* Use the last TLB for temporary access to LMB */
+#  define MICROBLAZE_LMB_TLB_ID 63
 
 /*
  * TLB entries are defined by a "high" tag portion and a "low" data
diff --git a/arch/microblaze/kernel/early_printk.c b/arch/microblaze/kernel/early_printk.c
index 742c247..ec48587 100644
--- a/arch/microblaze/kernel/early_printk.c
+++ b/arch/microblaze/kernel/early_printk.c
@@ -175,6 +175,20 @@
 								base_addr);
 	base_addr = (u32) ioremap(base_addr, PAGE_SIZE);
 	printk(KERN_CONT "0x%x\n", base_addr);
+
+	/*
+	 * Early console is on the top of skipped TLB entries
+	 * decrease tlb_skip size ensure that hardcoded TLB entry will be
+	 * used by generic algorithm
+	 * FIXME check if early console mapping is on the top by rereading
+	 * TLB entry and compare baseaddr
+	 *  mts  rtlbx, (tlb_skip - 1)
+	 *  nop
+	 *  mfs  rX, rtlblo
+	 *  nop
+	 *  cmp rX, orig_base_addr
+	 */
+	tlb_skip -= 1;
 }
 
 void __init disable_early_printk(void)
diff --git a/arch/microblaze/kernel/head.S b/arch/microblaze/kernel/head.S
index 49dd48f..98b17f9 100644
--- a/arch/microblaze/kernel/head.S
+++ b/arch/microblaze/kernel/head.S
@@ -149,6 +149,7 @@
 _invalidate:
 	mts	rtlbx, r3
 	mts	rtlbhi, r0			/* flush: ensure V is clear   */
+	mts	rtlblo, r0
 	bgtid	r3, _invalidate		/* loop for all entries       */
 	addik	r3, r3, -1
 	/* sync */
@@ -224,8 +225,14 @@
 	andi	r4,r4,0xfffffc00	/* Mask off the real page number */
 	ori	r4,r4,(TLB_WR | TLB_EX)	/* Set the write and execute bits */
 
-	/* TLB0 can be zeroes that's why we not setup it */
-	beqi	r9, jump_over
+	/*
+	 * TLB0 is always used - check if is not zero (r9 stores TLB0 value)
+	 * if is use TLB1 value and clear it (r10 stores TLB1 value)
+	 */
+	bnei	r9, tlb0_not_zero
+	add	r9, r10, r0
+	add	r10, r0, r0
+tlb0_not_zero:
 
 	/* look at the code below */
 	ori	r30, r0, 0x200
@@ -239,18 +246,21 @@
 	bneid	r29, 1f
 	addik	r30, r30, 0x80
 1:
-	ori r11, r30, 0
-
 	andi	r3,r3,0xfffffc00	/* Mask off the effective page number */
 	ori	r3,r3,(TLB_VALID)
-	or	r3, r3, r11
+	or	r3, r3, r30
 
-	mts     rtlbx,r0		/* TLB slow 0 */
+	/* Load tlb_skip size value which is index to first unused TLB entry */
+	lwi	r11, r0, TOPHYS(tlb_skip)
+	mts     rtlbx,r11		/* TLB slow 0 */
 
 	mts	rtlblo,r4		/* Load the data portion of the entry */
 	mts	rtlbhi,r3		/* Load the tag portion of the entry */
 
-jump_over:
+	/* Increase tlb_skip size */
+	addik	r11, r11, 1
+	swi	r11, r0, TOPHYS(tlb_skip)
+
 	/* TLB1 can be zeroes that's why we not setup it */
 	beqi	r10, jump_over2
 
@@ -266,27 +276,30 @@
 	bneid	r29, 1f
 	addik	r30, r30, 0x80
 1:
-	ori r12, r30, 0
-
 	addk	r4, r4, r9	/* previous addr + TLB0 size */
 	addk	r3, r3, r9
 
 	andi	r3,r3,0xfffffc00	/* Mask off the effective page number */
 	ori	r3,r3,(TLB_VALID)
-	or	r3, r3, r12
+	or	r3, r3, r30
 
-	ori	r6,r0,1			/* TLB slot 1 */
-	mts     rtlbx,r6
+	lwi	r11, r0, TOPHYS(tlb_skip)
+	mts     rtlbx, r11		/* r11 is used from TLB0 */
 
 	mts	rtlblo,r4		/* Load the data portion of the entry */
 	mts	rtlbhi,r3		/* Load the tag portion of the entry */
 
+	/* Increase tlb_skip size */
+	addik	r11, r11, 1
+	swi	r11, r0, TOPHYS(tlb_skip)
+
 jump_over2:
 	/*
 	 * Load a TLB entry for LMB, since we need access to
 	 * the exception vectors, using a 4k real==virtual mapping.
 	 */
-	ori	r6,r0,3			/* TLB slot 3 */
+	/* Use temporary TLB_ID for LMB - clear this temporary mapping later */
+	ori	r6, r0, MICROBLAZE_LMB_TLB_ID
 	mts     rtlbx,r6
 
 	ori	r4,r0,(TLB_WR | TLB_EX)
@@ -355,8 +368,7 @@
 
 	/* Load up the kernel context */
 kernel_load_context:
-	# Keep entry 0 and 1 valid. Entry 3 mapped to LMB can go away.
-	ori	r5,r0,3
+	ori	r5, r0, MICROBLAZE_LMB_TLB_ID
 	mts     rtlbx,r5
 	nop
 	mts	rtlbhi,r0
diff --git a/arch/microblaze/kernel/hw_exception_handler.S b/arch/microblaze/kernel/hw_exception_handler.S
index b7249f4..aa510f4 100644
--- a/arch/microblaze/kernel/hw_exception_handler.S
+++ b/arch/microblaze/kernel/hw_exception_handler.S
@@ -820,9 +820,15 @@
  *	Upon exit, we reload everything and RFI.
  * A common place to load the TLB.
  */
+.section .data
+.align 4
+.global tlb_skip
+	tlb_skip:
+		.long	MICROBLAZE_TLB_SKIP
 	tlb_index:
 		/* MS: storing last used tlb index */
-		.long	(MICROBLAZE_TLB_SKIP - 1)
+		.long	MICROBLAZE_TLB_SIZE/2
+.previous
 	finish_tlb_load:
 		/* MS: load the last used TLB index. */
 		lwi	r5, r0, TOPHYS(tlb_index)
@@ -833,7 +839,7 @@
 		ori	r6, r0, 1
 		cmp	r31, r5, r6
 		blti	r31, ex12
-		addik	r5, r6, MICROBLAZE_TLB_SKIP - 1
+		lwi	r5, r0, TOPHYS(tlb_skip)
 	ex12:
 		/* MS: save back current TLB index */
 		swi	r5, r0, TOPHYS(tlb_index)
diff --git a/arch/microblaze/kernel/misc.S b/arch/microblaze/kernel/misc.S
index c9090d7..1dafdde 100644
--- a/arch/microblaze/kernel/misc.S
+++ b/arch/microblaze/kernel/misc.S
@@ -29,16 +29,16 @@
 .type  _tlbia, @function
 .align 4;
 _tlbia:
-	addik	r12, r0, MICROBLAZE_TLB_SIZE - 1 /* flush all entries (63 - 3) */
+	lwi	r12, r0, tlb_skip;
 	/* isync */
 _tlbia_1:
 	mts	rtlbx, r12
 	nop
 	mts	rtlbhi, r0 /* flush: ensure V is clear */
 	nop
-	addik	r11, r12, -MICROBLAZE_TLB_SKIP
+	rsubi	r11, r12, MICROBLAZE_TLB_SIZE - 1
 	bneid	r11, _tlbia_1 /* loop for all entries */
-	addik	r12, r12, -1
+	addik	r12, r12, 1
 	/* sync */
 	rtsd	r15, 8
 	nop
@@ -75,7 +75,7 @@
 	 * Load a TLB entry for the UART, so that microblaze_progress() can use
 	 * the UARTs nice and early.  We use a 4k real==virtual mapping.
 	 */
-	ori	r4, r0, 63
+	lwi	r4, r0, tlb_skip
 	mts	rtlbx, r4 /* TLB slot 63 */
 
 	or	r4,r5,r0
@@ -89,6 +89,11 @@
 	nop
 	mts	rtlbhi,r5 /* Load the tag portion of the entry */
 	nop
+
+	lwi	r5, r0, tlb_skip
+	addik	r5, r5, 1
+	swi	r5, r0, tlb_skip
+
 	rtsd	r15, 8
 	nop
 
diff --git a/arch/microblaze/kernel/setup.c b/arch/microblaze/kernel/setup.c
index a1fa2a5..e4f5956 100644
--- a/arch/microblaze/kernel/setup.c
+++ b/arch/microblaze/kernel/setup.c
@@ -208,6 +208,19 @@
 	return of_debugfs_root == NULL;
 }
 arch_initcall(microblaze_debugfs_init);
+
+static int __init debugfs_tlb(void)
+{
+	struct dentry *d;
+
+	if (!of_debugfs_root)
+		return -ENODEV;
+
+	d = debugfs_create_u32("tlb_skip", S_IRUGO, of_debugfs_root, &tlb_skip);
+	if (!d)
+		return -ENOMEM;
+}
+device_initcall(debugfs_tlb);
 #endif
 
 static int dflt_bus_notify(struct notifier_block *nb,