Implement slbia as a dTLB/iTLB flush

Slbia (with IH=7) is used in the Linux kernel to flush the ERATs
(our iTLB/dTLB), so make it do that.

This moves the logic to work out whether to flush a single entry
or the whole TLB from dcache and icache into mmu.  We now invalidate
all dTLB and iTLB entries when the AP (actual pagesize) field of
RB is non-zero on a tlbie[l], as well as when IS is non-zero.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>

common.vhdl

@@ -227,6 +227,7 @@ package common is
 	valid : std_ulogic;
         op : insn_type_t;                               -- what ld/st or m[tf]spr or TLB op to do
         nia : std_ulogic_vector(63 downto 0);
+        insn : std_ulogic_vector(31 downto 0);
 	addr1 : std_ulogic_vector(63 downto 0);
 	addr2 : std_ulogic_vector(63 downto 0);
 	data : std_ulogic_vector(63 downto 0);		-- data to write, unused for read
@@ -242,12 +243,11 @@ package common is
         rc : std_ulogic;                                -- set for stcx.
         virt_mode : std_ulogic;                         -- do translation through TLB
         priv_mode : std_ulogic;                         -- privileged mode (MSR[PR] = 0)
-        spr_num : spr_num_t;                            -- SPR number for mfspr/mtspr
     end record;
     constant Execute1ToLoadstore1Init : Execute1ToLoadstore1Type := (valid => '0', op => OP_ILLEGAL, ci => '0', byte_reverse => '0',
                                                                      sign_extend => '0', update => '0', xerc => xerc_init,
                                                                      reserve => '0', rc => '0', virt_mode => '0', priv_mode => '0',
-                                                                     spr_num => 0, others => (others => '0'));
+                                                                     others => (others => '0'));
 
     type Loadstore1ToExecute1Type is record
         exception : std_ulogic;
@@ -283,6 +283,7 @@ package common is
     type Loadstore1ToMmuType is record
         valid : std_ulogic;
         tlbie : std_ulogic;
+        slbia : std_ulogic;
         mtspr : std_ulogic;
         iside : std_ulogic;
         load  : std_ulogic;
@@ -305,6 +306,7 @@ package common is
     type MmuToDcacheType is record
         valid : std_ulogic;
         tlbie : std_ulogic;
+        doall : std_ulogic;
         tlbld : std_ulogic;
         addr  : std_ulogic_vector(63 downto 0);
         pte   : std_ulogic_vector(63 downto 0);
@@ -320,6 +322,7 @@ package common is
     type MmuToIcacheType is record
         tlbld : std_ulogic;
         tlbie : std_ulogic;
+        doall : std_ulogic;
         addr  : std_ulogic_vector(63 downto 0);
         pte   : std_ulogic_vector(63 downto 0);
     end record;

dcache.vhdl

@@ -210,6 +210,7 @@ architecture rtl of dcache is
     type reg_stage_0_t is record
         req   : Loadstore1ToDcacheType;
         tlbie : std_ulogic;
+        doall : std_ulogic;
         tlbld : std_ulogic;
         mmu_req : std_ulogic;   -- indicates source of request
     end record;
@@ -456,11 +457,13 @@ begin
                     r0.req.data <= m_in.pte;
                     r0.req.byte_sel <= (others => '1');
                     r0.tlbie <= m_in.tlbie;
+                    r0.doall <= m_in.doall;
                     r0.tlbld <= m_in.tlbld;
                     r0.mmu_req <= '1';
                 else
                     r0.req <= d_in;
                     r0.tlbie <= '0';
+                    r0.doall <= '0';
                     r0.tlbld <= '0';
                     r0.mmu_req <= '0';
                 end if;
@@ -572,7 +575,6 @@ begin
 
     tlb_update : process(clk)
         variable tlbie : std_ulogic;
-        variable tlbia : std_ulogic;
         variable tlbwe : std_ulogic;
         variable repl_way : tlb_way_t;
         variable eatag : tlb_tag_t;
@@ -580,17 +582,9 @@ begin
         variable pteset : tlb_way_ptes_t;
     begin
         if rising_edge(clk) then
-            tlbie := '0';
+            tlbie := r0_valid and r0.tlbie;
-            tlbia := '0';
             tlbwe := r0_valid and r0.tlbld;
-            if r0_valid = '1' and r0.tlbie = '1' then
+            if rst = '1' or (tlbie = '1' and r0.doall = '1') then
-                if r0.req.addr(11 downto 10) /= "00" then
-                    tlbia := '1';
-                else
-                    tlbie := '1';
-                end if;
-            end if;
-            if rst = '1' or tlbia = '1' then
                 -- clear all valid bits at once
                 for i in tlb_index_t loop
                     dtlb_valids(i) <= (others => '0');

decode1.vhdl

@@ -282,6 +282,7 @@ architecture behaviour of decode1 is
 		2#0010111010#  =>       (ALU,    OP_PRTY,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is8B, '0', '0', '0', '0', '0', '0', NONE, '0', '0'), -- prtyd
 		2#0010011010#  =>       (ALU,    OP_PRTY,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is4B, '0', '0', '0', '0', '0', '0', NONE, '0', '0'), -- prtyw
 		-- 2#0010000000# setb
+                2#0111110010#  =>       (LDST,   OP_TLBIE,     NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0'), -- slbia
 		2#0000011011#  =>       (ALU,    OP_SHL,       NONE,       RB,          RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0'), -- sld
 		2#0000011000#  =>       (ALU,    OP_SHL,       NONE,       RB,          RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '0', RC,   '0', '0'), -- slw
 		2#1100011010#  =>       (ALU,    OP_SHR,       NONE,       RB,          RS,   RA,   '0', '0', '0', '0', ZERO, '1', NONE, '0', '0', '0', '0', '0', '1', RC,   '0', '0'), -- srad

execute1.vhdl

@@ -1023,7 +1023,7 @@ begin
         lv.xerc := v.e.xerc;
         lv.reserve := e_in.reserve;
         lv.rc := e_in.rc;
-        lv.spr_num := decode_spr_num(e_in.insn);
+        lv.insn := e_in.insn;
         -- decode l*cix and st*cix instructions here
         if e_in.insn(31 downto 26) = "011111" and e_in.insn(10 downto 9) = "11" and
             e_in.insn(5 downto 1) = "10101" then

icache.vhdl

@@ -435,32 +435,19 @@ begin
 
     -- iTLB update
     itlb_update: process(clk)
-        variable tlbie : std_ulogic;
-        variable tlbia : std_ulogic;
-        variable tlbwe : std_ulogic;
         variable wr_index : tlb_index_t;
     begin
         if rising_edge(clk) then
-            tlbie := '0';
-            tlbia := '0';
-            tlbwe := m_in.tlbld;
-            if m_in.tlbie = '1' then
-                if m_in.addr(11 downto 10) /= "00" then
-                    tlbia := '1';
-                else
-                    tlbie := '1';
-                end if;
-            end if;
             wr_index := hash_ea(m_in.addr);
-            if rst = '1' or tlbia = '1' then
+            if rst = '1' or (m_in.tlbie = '1' and m_in.doall = '1') then
                 -- clear all valid bits
                 for i in tlb_index_t loop
                     itlb_valids(i) <= '0';
                 end loop;
-            elsif tlbie = '1' then
+            elsif m_in.tlbie = '1' then
                 -- clear entry regardless of hit or miss
                 itlb_valids(wr_index) <= '0';
-            elsif tlbwe = '1' then
+            elsif m_in.tlbld = '1' then
                 itlb_tags(wr_index) <= m_in.addr(63 downto TLB_LG_PGSZ + TLB_BITS);
                 itlb_ptes(wr_index) <= m_in.pte;
                 itlb_valids(wr_index) <= '1';

loadstore1.vhdl

@@ -166,7 +166,7 @@ begin
         mfspr := '0';
         mmu_mtspr := '0';
         itlb_fault := '0';
-        sprn := std_ulogic_vector(to_unsigned(l_in.spr_num, 10));
+        sprn := std_ulogic_vector(to_unsigned(decode_spr_num(l_in.insn), 10));
         sprval := (others => '0');      -- avoid inferred latches
         exception := '0';
         dsisr := (others => '0');
@@ -468,6 +468,7 @@ begin
         m_out.mtspr <= mmu_mtspr;
         m_out.sprn <= sprn(3 downto 0);
         m_out.addr <= addr;
+        m_out.slbia <= l_in.insn(7);
         m_out.rs <= l_in.data;
 
         -- Update outputs to writeback

mmu.vhdl

@@ -168,6 +168,7 @@ begin
         variable tlb_load : std_ulogic;
         variable itlb_load : std_ulogic;
         variable tlbie_req : std_ulogic;
+        variable inval_all : std_ulogic;
         variable rts : unsigned(5 downto 0);
         variable mbits : unsigned(5 downto 0);
         variable pgtable_addr : std_ulogic_vector(63 downto 0);
@@ -191,6 +192,7 @@ begin
         tlb_load := '0';
         itlb_load := '0';
         tlbie_req := '0';
+        inval_all := '0';
 
         -- Radix tree data structures in memory are big-endian,
         -- so we need to byte-swap them
@@ -217,6 +219,10 @@ begin
                 if l_in.tlbie = '1' then
                     dcreq := '1';
                     tlbie_req := '1';
+                    -- Invalidate all iTLB/dTLB entries for tlbie with
+                    -- RB[IS] != 0 or RB[AP] != 0, or for slbia
+                    inval_all := l_in.slbia or l_in.addr(11) or l_in.addr(10) or
+                                 l_in.addr(7) or l_in.addr(6) or l_in.addr(5);
                     v.state := TLB_WAIT;
                 else
                     v.valid := '1';
@@ -356,12 +362,14 @@ begin
 
         d_out.valid <= dcreq;
         d_out.tlbie <= tlbie_req;
+        d_out.doall <= inval_all;
         d_out.tlbld <= tlb_load;
         d_out.addr <= addr;
         d_out.pte <= tlb_data;
 
         i_out.tlbld <= itlb_load;
         i_out.tlbie <= tlbie_req;
+        i_out.doall <= inval_all;
         i_out.addr <= addr;
         i_out.pte <= tlb_data;