diff options
Diffstat (limited to 'arch/sparc64/kernel/entry.S')
| -rw-r--r-- | arch/sparc64/kernel/entry.S | 262 |
1 files changed, 47 insertions, 215 deletions
diff --git a/arch/sparc64/kernel/entry.S b/arch/sparc64/kernel/entry.S index b4834952785..11a848402fb 100644 --- a/arch/sparc64/kernel/entry.S +++ b/arch/sparc64/kernel/entry.S @@ -30,163 +30,10 @@ .text .align 32 - .globl sparc64_vpte_patchme1 - .globl sparc64_vpte_patchme2 -/* - * On a second level vpte miss, check whether the original fault is to the OBP - * range (note that this is only possible for instruction miss, data misses to - * obp range do not use vpte). If so, go back directly to the faulting address. - * This is because we want to read the tpc, otherwise we have no way of knowing - * the 8k aligned faulting address if we are using >8k kernel pagesize. This - * also ensures no vpte range addresses are dropped into tlb while obp is - * executing (see inherit_locked_prom_mappings() rant). - */ -sparc64_vpte_nucleus: - /* Note that kvmap below has verified that the address is - * in the range MODULES_VADDR --> VMALLOC_END already. So - * here we need only check if it is an OBP address or not. - */ - sethi %hi(LOW_OBP_ADDRESS), %g5 - cmp %g4, %g5 - blu,pn %xcc, sparc64_vpte_patchme1 - mov 0x1, %g5 - sllx %g5, 32, %g5 - cmp %g4, %g5 - blu,pn %xcc, obp_iaddr_patch - nop - - /* These two instructions are patched by paginig_init(). */ -sparc64_vpte_patchme1: - sethi %hi(0), %g5 -sparc64_vpte_patchme2: - or %g5, %lo(0), %g5 - - /* With kernel PGD in %g5, branch back into dtlb_backend. */ - ba,pt %xcc, sparc64_kpte_continue - andn %g1, 0x3, %g1 /* Finish PMD offset adjustment. */ - -vpte_noent: - /* Restore previous TAG_ACCESS, %g5 is zero, and we will - * skip over the trap instruction so that the top level - * TLB miss handler will thing this %g5 value is just an - * invalid PTE, thus branching to full fault processing. - */ - mov TLB_SFSR, %g1 - stxa %g4, [%g1 + %g1] ASI_DMMU - done - - .globl obp_iaddr_patch -obp_iaddr_patch: - /* These two instructions patched by inherit_prom_mappings(). */ - sethi %hi(0), %g5 - or %g5, %lo(0), %g5 - - /* Behave as if we are at TL0. */ - wrpr %g0, 1, %tl - rdpr %tpc, %g4 /* Find original faulting iaddr */ - srlx %g4, 13, %g4 /* Throw out context bits */ - sllx %g4, 13, %g4 /* g4 has vpn + ctx0 now */ - - /* Restore previous TAG_ACCESS. */ - mov TLB_SFSR, %g1 - stxa %g4, [%g1 + %g1] ASI_IMMU - - /* Get PMD offset. */ - srlx %g4, 23, %g6 - and %g6, 0x7ff, %g6 - sllx %g6, 2, %g6 - - /* Load PMD, is it valid? */ - lduwa [%g5 + %g6] ASI_PHYS_USE_EC, %g5 - brz,pn %g5, longpath - sllx %g5, 11, %g5 - - /* Get PTE offset. */ - srlx %g4, 13, %g6 - and %g6, 0x3ff, %g6 - sllx %g6, 3, %g6 - - /* Load PTE. */ - ldxa [%g5 + %g6] ASI_PHYS_USE_EC, %g5 - brgez,pn %g5, longpath - nop - - /* TLB load and return from trap. */ - stxa %g5, [%g0] ASI_ITLB_DATA_IN - retry - - .globl obp_daddr_patch -obp_daddr_patch: - /* These two instructions patched by inherit_prom_mappings(). */ - sethi %hi(0), %g5 - or %g5, %lo(0), %g5 - - /* Get PMD offset. */ - srlx %g4, 23, %g6 - and %g6, 0x7ff, %g6 - sllx %g6, 2, %g6 - - /* Load PMD, is it valid? */ - lduwa [%g5 + %g6] ASI_PHYS_USE_EC, %g5 - brz,pn %g5, longpath - sllx %g5, 11, %g5 - - /* Get PTE offset. */ - srlx %g4, 13, %g6 - and %g6, 0x3ff, %g6 - sllx %g6, 3, %g6 - - /* Load PTE. */ - ldxa [%g5 + %g6] ASI_PHYS_USE_EC, %g5 - brgez,pn %g5, longpath - nop - - /* TLB load and return from trap. */ - stxa %g5, [%g0] ASI_DTLB_DATA_IN - retry - -/* - * On a first level data miss, check whether this is to the OBP range (note - * that such accesses can be made by prom, as well as by kernel using - * prom_getproperty on "address"), and if so, do not use vpte access ... - * rather, use information saved during inherit_prom_mappings() using 8k - * pagesize. - */ - .align 32 -kvmap: - sethi %hi(MODULES_VADDR), %g5 - cmp %g4, %g5 - blu,pn %xcc, longpath - mov (VMALLOC_END >> 24), %g5 - sllx %g5, 24, %g5 - cmp %g4, %g5 - bgeu,pn %xcc, longpath - nop - -kvmap_check_obp: - sethi %hi(LOW_OBP_ADDRESS), %g5 - cmp %g4, %g5 - blu,pn %xcc, kvmap_vmalloc_addr - mov 0x1, %g5 - sllx %g5, 32, %g5 - cmp %g4, %g5 - blu,pn %xcc, obp_daddr_patch - nop - -kvmap_vmalloc_addr: - /* If we get here, a vmalloc addr was accessed, load kernel VPTE. */ - ldxa [%g3 + %g6] ASI_N, %g5 - brgez,pn %g5, longpath - nop - - /* PTE is valid, load into TLB and return from trap. */ - stxa %g5, [%g0] ASI_DTLB_DATA_IN ! Reload TLB - retry - /* This is trivial with the new code... */ .globl do_fpdis do_fpdis: - sethi %hi(TSTATE_PEF), %g4 ! IEU0 + sethi %hi(TSTATE_PEF), %g4 rdpr %tstate, %g5 andcc %g5, %g4, %g0 be,pt %xcc, 1f @@ -203,18 +50,18 @@ do_fpdis: add %g0, %g0, %g0 ba,a,pt %xcc, rtrap_clr_l6 -1: ldub [%g6 + TI_FPSAVED], %g5 ! Load Group - wr %g0, FPRS_FEF, %fprs ! LSU Group+4bubbles - andcc %g5, FPRS_FEF, %g0 ! IEU1 Group - be,a,pt %icc, 1f ! CTI - clr %g7 ! IEU0 - ldx [%g6 + TI_GSR], %g7 ! Load Group -1: andcc %g5, FPRS_DL, %g0 ! IEU1 - bne,pn %icc, 2f ! CTI - fzero %f0 ! FPA - andcc %g5, FPRS_DU, %g0 ! IEU1 Group - bne,pn %icc, 1f ! CTI - fzero %f2 ! FPA +1: ldub [%g6 + TI_FPSAVED], %g5 + wr %g0, FPRS_FEF, %fprs + andcc %g5, FPRS_FEF, %g0 + be,a,pt %icc, 1f + clr %g7 + ldx [%g6 + TI_GSR], %g7 +1: andcc %g5, FPRS_DL, %g0 + bne,pn %icc, 2f + fzero %f0 + andcc %g5, FPRS_DU, %g0 + bne,pn %icc, 1f + fzero %f2 faddd %f0, %f2, %f4 fmuld %f0, %f2, %f6 faddd %f0, %f2, %f8 @@ -250,15 +97,17 @@ do_fpdis: faddd %f0, %f2, %f4 fmuld %f0, %f2, %f6 ldxa [%g3] ASI_DMMU, %g5 -cplus_fptrap_insn_1: - sethi %hi(0), %g2 + sethi %hi(sparc64_kern_sec_context), %g2 + ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2 stxa %g2, [%g3] ASI_DMMU membar #Sync add %g6, TI_FPREGS + 0xc0, %g2 faddd %f0, %f2, %f8 fmuld %f0, %f2, %f10 - ldda [%g1] ASI_BLK_S, %f32 ! grrr, where is ASI_BLK_NUCLEUS 8-( + membar #Sync + ldda [%g1] ASI_BLK_S, %f32 ldda [%g2] ASI_BLK_S, %f48 + membar #Sync faddd %f0, %f2, %f12 fmuld %f0, %f2, %f14 faddd %f0, %f2, %f16 @@ -269,7 +118,6 @@ cplus_fptrap_insn_1: fmuld %f0, %f2, %f26 faddd %f0, %f2, %f28 fmuld %f0, %f2, %f30 - membar #Sync b,pt %xcc, fpdis_exit nop 2: andcc %g5, FPRS_DU, %g0 @@ -279,15 +127,17 @@ cplus_fptrap_insn_1: fzero %f34 ldxa [%g3] ASI_DMMU, %g5 add %g6, TI_FPREGS, %g1 -cplus_fptrap_insn_2: - sethi %hi(0), %g2 + sethi %hi(sparc64_kern_sec_context), %g2 + ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2 stxa %g2, [%g3] ASI_DMMU membar #Sync add %g6, TI_FPREGS + 0x40, %g2 faddd %f32, %f34, %f36 fmuld %f32, %f34, %f38 - ldda [%g1] ASI_BLK_S, %f0 ! grrr, where is ASI_BLK_NUCLEUS 8-( + membar #Sync + ldda [%g1] ASI_BLK_S, %f0 ldda [%g2] ASI_BLK_S, %f16 + membar #Sync faddd %f32, %f34, %f40 fmuld %f32, %f34, %f42 faddd %f32, %f34, %f44 @@ -300,18 +150,18 @@ cplus_fptrap_insn_2: fmuld %f32, %f34, %f58 faddd %f32, %f34, %f60 fmuld %f32, %f34, %f62 - membar #Sync ba,pt %xcc, fpdis_exit nop 3: mov SECONDARY_CONTEXT, %g3 add %g6, TI_FPREGS, %g1 ldxa [%g3] ASI_DMMU, %g5 -cplus_fptrap_insn_3: - sethi %hi(0), %g2 + sethi %hi(sparc64_kern_sec_context), %g2 + ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2 stxa %g2, [%g3] ASI_DMMU membar #Sync mov 0x40, %g2 - ldda [%g1] ASI_BLK_S, %f0 ! grrr, where is ASI_BLK_NUCLEUS 8-( + membar #Sync + ldda [%g1] ASI_BLK_S, %f0 ldda [%g1 + %g2] ASI_BLK_S, %f16 add %g1, 0x80, %g1 ldda [%g1] ASI_BLK_S, %f32 @@ -472,8 +322,8 @@ do_fptrap_after_fsr: stx %g3, [%g6 + TI_GSR] mov SECONDARY_CONTEXT, %g3 ldxa [%g3] ASI_DMMU, %g5 -cplus_fptrap_insn_4: - sethi %hi(0), %g2 + sethi %hi(sparc64_kern_sec_context), %g2 + ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2 stxa %g2, [%g3] ASI_DMMU membar #Sync add %g6, TI_FPREGS, %g2 @@ -494,45 +344,17 @@ cplus_fptrap_insn_4: ba,pt %xcc, etrap wr %g0, 0, %fprs -cplus_fptrap_1: - sethi %hi(CTX_CHEETAH_PLUS_CTX0), %g2 - - .globl cheetah_plus_patch_fpdis -cheetah_plus_patch_fpdis: - /* We configure the dTLB512_0 for 4MB pages and the - * dTLB512_1 for 8K pages when in context zero. - */ - sethi %hi(cplus_fptrap_1), %o0 - lduw [%o0 + %lo(cplus_fptrap_1)], %o1 - - set cplus_fptrap_insn_1, %o2 - stw %o1, [%o2] - flush %o2 - set cplus_fptrap_insn_2, %o2 - stw %o1, [%o2] - flush %o2 - set cplus_fptrap_insn_3, %o2 - stw %o1, [%o2] - flush %o2 - set cplus_fptrap_insn_4, %o2 - stw %o1, [%o2] - flush %o2 - - retl - nop - /* The registers for cross calls will be: * * DATA 0: [low 32-bits] Address of function to call, jmp to this * [high 32-bits] MMU Context Argument 0, place in %g5 - * DATA 1: Address Argument 1, place in %g6 + * DATA 1: Address Argument 1, place in %g1 * DATA 2: Address Argument 2, place in %g7 * * With this method we can do most of the cross-call tlb/cache * flushing very quickly. * - * Current CPU's IRQ worklist table is locked into %g1, - * don't touch. + * Current CPU's IRQ worklist table is locked into %g6, don't touch. */ .text .align 32 @@ -1006,13 +828,14 @@ cheetah_plus_dcpe_trap_vector: nop do_cheetah_plus_data_parity: - ba,pt %xcc, etrap + rdpr %pil, %g2 + wrpr %g0, 15, %pil + ba,pt %xcc, etrap_irq rd %pc, %g7 mov 0x0, %o0 call cheetah_plus_parity_error add %sp, PTREGS_OFF, %o1 - ba,pt %xcc, rtrap - clr %l6 + ba,a,pt %xcc, rtrap_irq cheetah_plus_dcpe_trap_vector_tl1: membar #Sync @@ -1036,13 +859,14 @@ cheetah_plus_icpe_trap_vector: nop do_cheetah_plus_insn_parity: - ba,pt %xcc, etrap + rdpr %pil, %g2 + wrpr %g0, 15, %pil + ba,pt %xcc, etrap_irq rd %pc, %g7 mov 0x1, %o0 call cheetah_plus_parity_error add %sp, PTREGS_OFF, %o1 - ba,pt %xcc, rtrap - clr %l6 + ba,a,pt %xcc, rtrap_irq cheetah_plus_icpe_trap_vector_tl1: membar #Sync @@ -1075,6 +899,10 @@ do_dcpe_tl1: nop wrpr %g1, %tl ! Restore original trap level do_dcpe_tl1_nonfatal: /* Ok we may use interrupt globals safely. */ + sethi %hi(dcache_parity_tl1_occurred), %g2 + lduw [%g2 + %lo(dcache_parity_tl1_occurred)], %g1 + add %g1, 1, %g1 + stw %g1, [%g2 + %lo(dcache_parity_tl1_occurred)] /* Reset D-cache parity */ sethi %hi(1 << 16), %g1 ! D-cache size mov (1 << 5), %g2 ! D-cache line size @@ -1121,6 +949,10 @@ do_icpe_tl1: nop wrpr %g1, %tl ! Restore original trap level do_icpe_tl1_nonfatal: /* Ok we may use interrupt globals safely. */ + sethi %hi(icache_parity_tl1_occurred), %g2 + lduw [%g2 + %lo(icache_parity_tl1_occurred)], %g1 + add %g1, 1, %g1 + stw %g1, [%g2 + %lo(icache_parity_tl1_occurred)] /* Flush I-cache */ sethi %hi(1 << 15), %g1 ! I-cache size mov (1 << 5), %g2 ! I-cache line size |