Blackfin arch: Faster C implementation of no-MPU CPLB handler

This is a mixture ofcMichael McTernan's patch and the existing cplb-mpu code.

We ditch the old cplb-nompu implementation, which is a good example of
why a good algorithm in a HLL is preferrable to a bad algorithm written in
assembly.  Rather than try to construct a table of all posible CPLBs and
search it, we just create a (smaller) table of memory regions and
their attributes.  Some of the data structures are now unified for both
the mpu and nompu cases.  A lot of needless complexity in cplbinit.c is
removed.

Further optimizations:
  * compile cplbmgr.c with a lot of -ffixed-reg options, and omit saving
    these registers on the stack when entering a CPLB exception.
  * lose cli/nop/nop/sti sequences for some workarounds - these don't
  * make
    sense in an exception context

Additional code unification should be possible after this.

[Mike Frysinger <vapier.adi@gmail.com>:
 - convert CPP if statements to C if statements
 - remove redundant statements
 - use a do...while loop rather than a for loop to get slightly better
   optimization and to avoid gcc "may be used uninitialized" warnings ...
   we know that the [id]cplb_nr_bounds variables will never be 0, so this
   is OK
 - the no-mpu code was the last user of MAX_MEM_SIZE and with that rewritten,
   we can punt it
 - add some BUG_ON() checks to make sure we dont overflow the small
   cplb_bounds array
 - add i/d cplb entries for the bootrom because there is functions/data in
   there we want to access
 - we do not need a NULL trailing entry as any time we access the bounds
   arrays, we use the nr_bounds variable
]

Signed-off-by: Michael McTernan <mmcternan@airvana.com>
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>

arch/blackfin/Kconfig

@@ -524,14 +524,6 @@ config MEM_SDGCTL
 	default 0x0
 endmenu
 
-config MAX_MEM_SIZE
-	int "Max SDRAM Memory Size in MBytes"
-	depends on !MPU
-	default 512
-	help
-	  This is the max memory size that the kernel will create CPLB
-	  tables for.  Your system will not be able to handle any more.
-
 #
 # Max & Min Speeds for various Chips
 #

arch/blackfin/include/asm/context.S

@@ -357,3 +357,42 @@
 	SYSCFG = [sp++];
 	csync;
 .endm
+
+.macro save_context_cplb
+	[--sp] = (R7:0, P5:0);
+	[--sp] = fp;
+
+	[--sp] = a0.x;
+	[--sp] = a0.w;
+	[--sp] = a1.x;
+	[--sp] = a1.w;
+
+	[--sp] = LC0;
+	[--sp] = LC1;
+	[--sp] = LT0;
+	[--sp] = LT1;
+	[--sp] = LB0;
+	[--sp] = LB1;
+
+	[--sp] = RETS;
+.endm
+
+.macro restore_context_cplb
+	RETS = [sp++];
+
+	LB1 = [sp++];
+	LB0 = [sp++];
+	LT1 = [sp++];
+	LT0 = [sp++];
+	LC1 = [sp++];
+	LC0 = [sp++];
+
+	a1.w = [sp++];
+	a1.x = [sp++];
+	a0.w = [sp++];
+	a0.x = [sp++];
+
+	fp = [sp++];
+
+	(R7:0, P5:0) = [SP++];
+.endm

arch/blackfin/include/asm/cplb.h

@@ -116,4 +116,8 @@
 #define CPLB_INOCACHE   	CPLB_USER_RD | CPLB_VALID
 #define CPLB_IDOCACHE   	CPLB_INOCACHE | CPLB_L1_CHBL
 
+#define FAULT_RW        (1 << 16)
+#define FAULT_USERSUPV  (1 << 17)
+#define FAULT_CPLBBITS  0x0000ffff
+
 #endif				/* _CPLB_H */

arch/blackfin/include/asm/cplbinit.h

@@ -32,96 +32,56 @@
 
 #include <asm/blackfin.h>
 #include <asm/cplb.h>
+#include <linux/threads.h>
 
-#ifdef CONFIG_MPU
-
-#include <asm/cplb-mpu.h>
-extern void bfin_icache_init(struct cplb_entry *icplb_tbl);
-extern void bfin_dcache_init(struct cplb_entry *icplb_tbl);
-
+#ifdef CONFIG_CPLB_SWITCH_TAB_L1
+# define PDT_ATTR __attribute__((l1_data))
 #else
+# define PDT_ATTR
+#endif
 
-#define INITIAL_T 0x1
-#define SWITCH_T  0x2
-#define I_CPLB    0x4
-#define D_CPLB    0x8
+struct cplb_entry {
+	unsigned long data, addr;
+};
 
-#define ASYNC_MEMORY_CPLB_COVERAGE  ((ASYNC_BANK0_SIZE + ASYNC_BANK1_SIZE + \
-				ASYNC_BANK2_SIZE + ASYNC_BANK3_SIZE) / SIZE_4M)
+struct cplb_boundary {
+	unsigned long eaddr; /* End of this region.  */
+	unsigned long data; /* CPLB data value.  */
+};
 
-#define CPLB_MEM CONFIG_MAX_MEM_SIZE
+extern struct cplb_boundary dcplb_bounds[];
+extern struct cplb_boundary icplb_bounds[];
+extern int dcplb_nr_bounds, icplb_nr_bounds;
 
-/*
-* Number of required data CPLB switchtable entries
-* MEMSIZE / 4 (we mostly install 4M page size CPLBs
-* approx 16 for smaller 1MB page size CPLBs for allignment purposes
-* 1 for L1 Data Memory
-* possibly 1 for L2 Data Memory
-* 1 for CONFIG_DEBUG_HUNT_FOR_ZERO
-* 1 for ASYNC Memory
-*/
-#define MAX_SWITCH_D_CPLBS (((CPLB_MEM / 4) + 16 + 1 + 1 + 1 \
-				 + ASYNC_MEMORY_CPLB_COVERAGE) * 2)
+extern struct cplb_entry dcplb_tbl[NR_CPUS][MAX_CPLBS];
+extern struct cplb_entry icplb_tbl[NR_CPUS][MAX_CPLBS];
+extern int first_switched_icplb;
+extern int first_switched_dcplb;
 
-/*
-* Number of required instruction CPLB switchtable entries
-* MEMSIZE / 4 (we mostly install 4M page size CPLBs
-* approx 12 for smaller 1MB page size CPLBs for allignment purposes
-* 1 for L1 Instruction Memory
-* possibly 1 for L2 Instruction Memory
-* 1 for CONFIG_DEBUG_HUNT_FOR_ZERO
-*/
-#define MAX_SWITCH_I_CPLBS (((CPLB_MEM / 4) + 12 + 1 + 1 + 1) * 2)
-
-/* Number of CPLB table entries, used for cplb-nompu. */
-#define CPLB_TBL_ENTRIES (16 * 4)
-
-enum {
-	ZERO_P, L1I_MEM, L1D_MEM, L2_MEM, SDRAM_KERN, SDRAM_RAM_MTD, SDRAM_DMAZ,
-	RES_MEM, ASYNC_MEM, OCB_ROM
-};
+extern int nr_dcplb_miss[], nr_icplb_miss[], nr_icplb_supv_miss[];
+extern int nr_dcplb_prot[], nr_cplb_flush[];
 
-struct cplb_desc {
-	u32 start; /* start address */
-	u32 end; /* end address */
-	u32 psize; /* prefered size if any otherwise 1MB or 4MB*/
-	u16 attr;/* attributes */
-	u16 i_conf;/* I-CPLB DATA */
-	u16 d_conf;/* D-CPLB DATA */
-	u16 valid;/* valid */
-	const s8 name[30];/* name */
-};
+#ifdef CONFIG_MPU
 
-struct cplb_tab {
-  u_long *tab;
-	u16 pos;
-	u16 size;
-};
+extern int first_mask_dcplb;
 
-extern u_long icplb_tables[NR_CPUS][CPLB_TBL_ENTRIES+1];
-extern u_long dcplb_tables[NR_CPUS][CPLB_TBL_ENTRIES+1];
+extern int page_mask_order;
+extern int page_mask_nelts;
 
-/* Till here we are discussing about the static memory management model.
- * However, the operating envoronments commonly define more CPLB
- * descriptors to cover the entire addressable memory than will fit into
- * the available on-chip 16 CPLB MMRs. When this happens, the below table
- * will be used which will hold all the potentially required CPLB descriptors
- *
- * This is how Page descriptor Table is implemented in uClinux/Blackfin.
- */
+extern unsigned long *current_rwx_mask[NR_CPUS];
 
-extern u_long ipdt_tables[NR_CPUS][MAX_SWITCH_I_CPLBS+1];
-extern u_long dpdt_tables[NR_CPUS][MAX_SWITCH_D_CPLBS+1];
-#ifdef CONFIG_CPLB_INFO
-extern u_long ipdt_swapcount_tables[NR_CPUS][MAX_SWITCH_I_CPLBS];
-extern u_long dpdt_swapcount_tables[NR_CPUS][MAX_SWITCH_D_CPLBS];
-#endif
-extern void bfin_icache_init(u_long icplbs[]);
-extern void bfin_dcache_init(u_long dcplbs[]);
+extern void flush_switched_cplbs(unsigned int);
+extern void set_mask_dcplbs(unsigned long *, unsigned int);
+
+extern void __noreturn panic_cplb_error(int seqstat, struct pt_regs *);
 
 #endif /* CONFIG_MPU */
 
+extern void bfin_icache_init(struct cplb_entry *icplb_tbl);
+extern void bfin_dcache_init(struct cplb_entry *icplb_tbl);
+
 #if defined(CONFIG_BFIN_DCACHE) || defined(CONFIG_BFIN_ICACHE)
+extern void generate_cplb_tables_all(void);
 extern void generate_cplb_tables_cpu(unsigned int cpu);
 #endif
 #endif

arch/blackfin/include/asm/entry.h

@@ -53,9 +53,11 @@
 /* This one pushes RETI without using CLI.  Interrupts are enabled.  */
 #define SAVE_CONTEXT_SYSCALL	save_context_syscall
 #define SAVE_CONTEXT		save_context_with_interrupts
+#define SAVE_CONTEXT_CPLB	save_context_cplb
 
 #define RESTORE_ALL_SYS		restore_context_no_interrupts
 #define RESTORE_CONTEXT		restore_context_with_interrupts
+#define RESTORE_CONTEXT_CPLB	restore_context_cplb
 
 #endif				/* __ASSEMBLY__ */
 #endif				/* __BFIN_ENTRY_H */

arch/blackfin/kernel/cplb-mpu/Makefile

@@ -3,3 +3,8 @@
 #
 
 obj-y := cplbinit.o cacheinit.o cplbmgr.o
+
+CFLAGS_cplbmgr.o := -ffixed-I0 -ffixed-I1 -ffixed-I2 -ffixed-I3 \
+		    -ffixed-L0 -ffixed-L1 -ffixed-L2 -ffixed-L3 \
+		    -ffixed-M0 -ffixed-M1 -ffixed-M2 -ffixed-M3 \
+		    -ffixed-B0 -ffixed-B1 -ffixed-B2 -ffixed-B3

arch/blackfin/kernel/cplb-mpu/cplbinit.c

@@ -107,3 +107,7 @@ void __init generate_cplb_tables_cpu(unsigned int cpu)
 	while (i_i < MAX_CPLBS)
 		icplb_tbl[cpu][i_i++].data = 0;
 }
+
+void generate_cplb_tables_all(void)
+{
+}

arch/blackfin/kernel/cplb-mpu/cplbmgr.c

@@ -25,8 +25,13 @@
 #include <asm/cplbinit.h>
 #include <asm/mmu_context.h>
 
-#define FAULT_RW	(1 << 16)
-#define FAULT_USERSUPV	(1 << 17)
+/*
+ * WARNING
+ *
+ * This file is compiled with certain -ffixed-reg options.  We have to
+ * make sure not to call any functions here that could clobber these
+ * registers.
+ */
 
 int page_mask_nelts;
 int page_mask_order;

arch/blackfin/kernel/cplb-nompu/Makefile

@@ -2,4 +2,9 @@
 # arch/blackfin/kernel/cplb-nompu/Makefile
 #
 
-obj-y := cplbinit.o cacheinit.o cplbhdlr.o cplbmgr.o
+obj-y := cplbinit.o cacheinit.o cplbmgr.o
+
+CFLAGS_cplbmgr.o := -ffixed-I0 -ffixed-I1 -ffixed-I2 -ffixed-I3 \
+		    -ffixed-L0 -ffixed-L1 -ffixed-L2 -ffixed-L3 \
+		    -ffixed-M0 -ffixed-M1 -ffixed-M2 -ffixed-M3 \
+		    -ffixed-B0 -ffixed-B1 -ffixed-B2 -ffixed-B3

arch/blackfin/kernel/cplb-nompu/cacheinit.c

@@ -25,19 +25,15 @@
 #include <asm/cplbinit.h>
 
 #if defined(CONFIG_BFIN_ICACHE)
-void __cpuinit bfin_icache_init(u_long icplb[])
+void __cpuinit bfin_icache_init(struct cplb_entry *icplb_tbl)
 {
-	unsigned long *table = icplb;
 	unsigned long ctrl;
 	int i;
 
+	SSYNC();
 	for (i = 0; i < MAX_CPLBS; i++) {
-		unsigned long addr = *table++;
-		unsigned long data = *table++;
-		if (addr == (unsigned long)-1)
-			break;
-		bfin_write32(ICPLB_ADDR0 + i * 4, addr);
-		bfin_write32(ICPLB_DATA0 + i * 4, data);
+		bfin_write32(ICPLB_ADDR0 + i * 4, icplb_tbl[i].addr);
+		bfin_write32(ICPLB_DATA0 + i * 4, icplb_tbl[i].data);
 	}
 	ctrl = bfin_read_IMEM_CONTROL();
 	ctrl |= IMC | ENICPLB;
@@ -47,24 +43,20 @@ void __cpuinit bfin_icache_init(u_long icplb[])
 #endif
 
 #if defined(CONFIG_BFIN_DCACHE)
-void __cpuinit bfin_dcache_init(u_long dcplb[])
+void __cpuinit bfin_dcache_init(struct cplb_entry *dcplb_tbl)
 {
-	unsigned long *table = dcplb;
 	unsigned long ctrl;
 	int i;
 
+	SSYNC();
 	for (i = 0; i < MAX_CPLBS; i++) {
-		unsigned long addr = *table++;
-		unsigned long data = *table++;
-		if (addr == (unsigned long)-1)
-			break;
-		bfin_write32(DCPLB_ADDR0 + i * 4, addr);
-		bfin_write32(DCPLB_DATA0 + i * 4, data);
+		bfin_write32(DCPLB_ADDR0 + i * 4, dcplb_tbl[i].addr);
+		bfin_write32(DCPLB_DATA0 + i * 4, dcplb_tbl[i].data);
 	}
+
 	ctrl = bfin_read_DMEM_CONTROL();
 	ctrl |= DMEM_CNTR;
 	bfin_write_DMEM_CONTROL(ctrl);
-
 	SSYNC();
 }
 #endif