diff options
author | Michael Neuling <mikey@neuling.org> | 2013-02-13 16:21:41 +0000 |
---|---|---|
committer | Benjamin Herrenschmidt <benh@kernel.crashing.org> | 2013-02-15 17:02:23 +1100 |
commit | 2b0a576d15e0e14751f00f9c87e46bad27f217e7 (patch) | |
tree | bd160f2f67cbae826fe0216bb9e217d9c3a290c8 /arch/powerpc | |
parent | bc2a9408fa65195288b41751016c36fd00a75a85 (diff) |
powerpc: Add new transactional memory state to the signal context
This adds the new transactional memory archtected state to the signal context
in both 32 and 64 bit.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Diffstat (limited to 'arch/powerpc')
-rw-r--r-- | arch/powerpc/include/asm/reg.h | 1 | ||||
-rw-r--r-- | arch/powerpc/kernel/signal.h | 8 | ||||
-rw-r--r-- | arch/powerpc/kernel/signal_32.c | 500 | ||||
-rw-r--r-- | arch/powerpc/kernel/signal_64.c | 337 |
4 files changed, 830 insertions, 16 deletions
diff --git a/arch/powerpc/include/asm/reg.h b/arch/powerpc/include/asm/reg.h index eee2a60994bf..7035e608f3fa 100644 --- a/arch/powerpc/include/asm/reg.h +++ b/arch/powerpc/include/asm/reg.h @@ -120,6 +120,7 @@ #define TM_CAUSE_FAC_UNAV 0xfa #define TM_CAUSE_SYSCALL 0xf9 /* Persistent */ #define TM_CAUSE_MISC 0xf6 +#define TM_CAUSE_SIGNAL 0xf4 #if defined(CONFIG_PPC_BOOK3S_64) #define MSR_64BIT MSR_SF diff --git a/arch/powerpc/kernel/signal.h b/arch/powerpc/kernel/signal.h index e00acb413934..ec84c901ceab 100644 --- a/arch/powerpc/kernel/signal.h +++ b/arch/powerpc/kernel/signal.h @@ -25,13 +25,21 @@ extern int handle_rt_signal32(unsigned long sig, struct k_sigaction *ka, extern unsigned long copy_fpr_to_user(void __user *to, struct task_struct *task); +extern unsigned long copy_transact_fpr_to_user(void __user *to, + struct task_struct *task); extern unsigned long copy_fpr_from_user(struct task_struct *task, void __user *from); +extern unsigned long copy_transact_fpr_from_user(struct task_struct *task, + void __user *from); #ifdef CONFIG_VSX extern unsigned long copy_vsx_to_user(void __user *to, struct task_struct *task); +extern unsigned long copy_transact_vsx_to_user(void __user *to, + struct task_struct *task); extern unsigned long copy_vsx_from_user(struct task_struct *task, void __user *from); +extern unsigned long copy_transact_vsx_from_user(struct task_struct *task, + void __user *from); #endif #ifdef CONFIG_PPC64 diff --git a/arch/powerpc/kernel/signal_32.c b/arch/powerpc/kernel/signal_32.c index 804e323c139d..e4a88d340de6 100644 --- a/arch/powerpc/kernel/signal_32.c +++ b/arch/powerpc/kernel/signal_32.c @@ -43,6 +43,7 @@ #include <asm/sigcontext.h> #include <asm/vdso.h> #include <asm/switch_to.h> +#include <asm/tm.h> #ifdef CONFIG_PPC64 #include "ppc32.h" #include <asm/unistd.h> @@ -293,6 +294,10 @@ long sys_sigaction(int sig, struct old_sigaction __user *act, struct sigframe { struct sigcontext sctx; /* the sigcontext */ struct mcontext mctx; /* all the register values */ +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + struct sigcontext sctx_transact; + struct mcontext mctx_transact; +#endif /* * Programs using the rs6000/xcoff abi can save up to 19 gp * regs and 18 fp regs below sp before decrementing it. @@ -321,6 +326,9 @@ struct rt_sigframe { struct siginfo info; #endif struct ucontext uc; +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + struct ucontext uc_transact; +#endif /* * Programs using the rs6000/xcoff abi can save up to 19 gp * regs and 18 fp regs below sp before decrementing it. @@ -381,6 +389,61 @@ unsigned long copy_vsx_from_user(struct task_struct *task, task->thread.fpr[i][TS_VSRLOWOFFSET] = buf[i]; return 0; } + +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +unsigned long copy_transact_fpr_to_user(void __user *to, + struct task_struct *task) +{ + double buf[ELF_NFPREG]; + int i; + + /* save FPR copy to local buffer then write to the thread_struct */ + for (i = 0; i < (ELF_NFPREG - 1) ; i++) + buf[i] = task->thread.TS_TRANS_FPR(i); + memcpy(&buf[i], &task->thread.transact_fpscr, sizeof(double)); + return __copy_to_user(to, buf, ELF_NFPREG * sizeof(double)); +} + +unsigned long copy_transact_fpr_from_user(struct task_struct *task, + void __user *from) +{ + double buf[ELF_NFPREG]; + int i; + + if (__copy_from_user(buf, from, ELF_NFPREG * sizeof(double))) + return 1; + for (i = 0; i < (ELF_NFPREG - 1) ; i++) + task->thread.TS_TRANS_FPR(i) = buf[i]; + memcpy(&task->thread.transact_fpscr, &buf[i], sizeof(double)); + + return 0; +} + +unsigned long copy_transact_vsx_to_user(void __user *to, + struct task_struct *task) +{ + double buf[ELF_NVSRHALFREG]; + int i; + + /* save FPR copy to local buffer then write to the thread_struct */ + for (i = 0; i < ELF_NVSRHALFREG; i++) + buf[i] = task->thread.transact_fpr[i][TS_VSRLOWOFFSET]; + return __copy_to_user(to, buf, ELF_NVSRHALFREG * sizeof(double)); +} + +unsigned long copy_transact_vsx_from_user(struct task_struct *task, + void __user *from) +{ + double buf[ELF_NVSRHALFREG]; + int i; + + if (__copy_from_user(buf, from, ELF_NVSRHALFREG * sizeof(double))) + return 1; + for (i = 0; i < ELF_NVSRHALFREG ; i++) + task->thread.transact_fpr[i][TS_VSRLOWOFFSET] = buf[i]; + return 0; +} +#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */ #else inline unsigned long copy_fpr_to_user(void __user *to, struct task_struct *task) @@ -395,6 +458,22 @@ inline unsigned long copy_fpr_from_user(struct task_struct *task, return __copy_from_user(task->thread.fpr, from, ELF_NFPREG * sizeof(double)); } + +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +inline unsigned long copy_transact_fpr_to_user(void __user *to, + struct task_struct *task) +{ + return __copy_to_user(to, task->thread.transact_fpr, + ELF_NFPREG * sizeof(double)); +} + +inline unsigned long copy_transact_fpr_from_user(struct task_struct *task, + void __user *from) +{ + return __copy_from_user(task->thread.transact_fpr, from, + ELF_NFPREG * sizeof(double)); +} +#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */ #endif /* @@ -483,6 +562,156 @@ static int save_user_regs(struct pt_regs *regs, struct mcontext __user *frame, return 0; } +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +/* + * Save the current user registers on the user stack. + * We only save the altivec/spe registers if the process has used + * altivec/spe instructions at some point. + * We also save the transactional registers to a second ucontext in the + * frame. + * + * See save_user_regs() and signal_64.c:setup_tm_sigcontexts(). + */ +static int save_tm_user_regs(struct pt_regs *regs, + struct mcontext __user *frame, + struct mcontext __user *tm_frame, int sigret) +{ + unsigned long msr = regs->msr; + + /* tm_reclaim rolls back all reg states, updating thread.ckpt_regs, + * thread.transact_fpr[], thread.transact_vr[], etc. + */ + tm_enable(); + tm_reclaim(¤t->thread, msr, TM_CAUSE_SIGNAL); + + /* Make sure floating point registers are stored in regs */ + flush_fp_to_thread(current); + + /* Save both sets of general registers */ + if (save_general_regs(¤t->thread.ckpt_regs, frame) + || save_general_regs(regs, tm_frame)) + return 1; + + /* Stash the top half of the 64bit MSR into the 32bit MSR word + * of the transactional mcontext. This way we have a backward-compatible + * MSR in the 'normal' (checkpointed) mcontext and additionally one can + * also look at what type of transaction (T or S) was active at the + * time of the signal. + */ + if (__put_user((msr >> 32), &tm_frame->mc_gregs[PT_MSR])) + return 1; + +#ifdef CONFIG_ALTIVEC + /* save altivec registers */ + if (current->thread.used_vr) { + flush_altivec_to_thread(current); + if (__copy_to_user(&frame->mc_vregs, current->thread.vr, + ELF_NVRREG * sizeof(vector128))) + return 1; + if (msr & MSR_VEC) { + if (__copy_to_user(&tm_frame->mc_vregs, + current->thread.transact_vr, + ELF_NVRREG * sizeof(vector128))) + return 1; + } else { + if (__copy_to_user(&tm_frame->mc_vregs, + current->thread.vr, + ELF_NVRREG * sizeof(vector128))) + return 1; + } + + /* set MSR_VEC in the saved MSR value to indicate that + * frame->mc_vregs contains valid data + */ + msr |= MSR_VEC; + } + + /* We always copy to/from vrsave, it's 0 if we don't have or don't + * use altivec. Since VSCR only contains 32 bits saved in the least + * significant bits of a vector, we "cheat" and stuff VRSAVE in the + * most significant bits of that same vector. --BenH + */ + if (__put_user(current->thread.vrsave, + (u32 __user *)&frame->mc_vregs[32])) + return 1; + if (msr & MSR_VEC) { + if (__put_user(current->thread.transact_vrsave, + (u32 __user *)&tm_frame->mc_vregs[32])) + return 1; + } else { + if (__put_user(current->thread.vrsave, + (u32 __user *)&tm_frame->mc_vregs[32])) + return 1; + } +#endif /* CONFIG_ALTIVEC */ + + if (copy_fpr_to_user(&frame->mc_fregs, current)) + return 1; + if (msr & MSR_FP) { + if (copy_transact_fpr_to_user(&tm_frame->mc_fregs, current)) + return 1; + } else { + if (copy_fpr_to_user(&tm_frame->mc_fregs, current)) + return 1; + } + +#ifdef CONFIG_VSX + /* + * Copy VSR 0-31 upper half from thread_struct to local + * buffer, then write that to userspace. Also set MSR_VSX in + * the saved MSR value to indicate that frame->mc_vregs + * contains valid data + */ + if (current->thread.used_vsr) { + __giveup_vsx(current); + if (copy_vsx_to_user(&frame->mc_vsregs, current)) + return 1; + if (msr & MSR_VSX) { + if (copy_transact_vsx_to_user(&tm_frame->mc_vsregs, + current)) + return 1; + } else { + if (copy_vsx_to_user(&tm_frame->mc_vsregs, current)) + return 1; + } + + msr |= MSR_VSX; + } +#endif /* CONFIG_VSX */ +#ifdef CONFIG_SPE + /* SPE regs are not checkpointed with TM, so this section is + * simply the same as in save_user_regs(). + */ + if (current->thread.used_spe) { + flush_spe_to_thread(current); + if (__copy_to_user(&frame->mc_vregs, current->thread.evr, + ELF_NEVRREG * sizeof(u32))) + return 1; + /* set MSR_SPE in the saved MSR value to indicate that + * frame->mc_vregs contains valid data */ + msr |= MSR_SPE; + } + + /* We always copy to/from spefscr */ + if (__put_user(current->thread.spefscr, (u32 __user *)&frame->mc_vregs + ELF_NEVRREG)) + return 1; +#endif /* CONFIG_SPE */ + + if (__put_user(msr, &frame->mc_gregs[PT_MSR])) + return 1; + if (sigret) { + /* Set up the sigreturn trampoline: li r0,sigret; sc */ + if (__put_user(0x38000000UL + sigret, &frame->tramp[0]) + || __put_user(0x44000002UL, &frame->tramp[1])) + return 1; + flush_icache_range((unsigned long) &frame->tramp[0], + (unsigned long) &frame->tramp[2]); + } + + return 0; +} +#endif + /* * Restore the current user register values from the user stack, * (except for MSR). @@ -588,6 +817,139 @@ static long restore_user_regs(struct pt_regs *regs, return 0; } +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +/* + * Restore the current user register values from the user stack, except for + * MSR, and recheckpoint the original checkpointed register state for processes + * in transactions. + */ +static long restore_tm_user_regs(struct pt_regs *regs, + struct mcontext __user *sr, + struct mcontext __user *tm_sr) +{ + long err; + unsigned long msr; +#ifdef CONFIG_VSX + int i; +#endif + + /* + * restore general registers but not including MSR or SOFTE. Also + * take care of keeping r2 (TLS) intact if not a signal. + * See comment in signal_64.c:restore_tm_sigcontexts(); + * TFHAR is restored from the checkpointed NIP; TEXASR and TFIAR + * were set by the signal delivery. + */ + err = restore_general_regs(regs, tm_sr); + err |= restore_general_regs(¤t->thread.ckpt_regs, sr); + + err |= __get_user(current->thread.tm_tfhar, &sr->mc_gregs[PT_NIP]); + + err |= __get_user(msr, &sr->mc_gregs[PT_MSR]); + if (err) + return 1; + + /* Restore the previous little-endian mode */ + regs->msr = (regs->msr & ~MSR_LE) | (msr & MSR_LE); + + /* + * Do this before updating the thread state in + * current->thread.fpr/vr/evr. That way, if we get preempted + * and another task grabs the FPU/Altivec/SPE, it won't be + * tempted to save the current CPU state into the thread_struct + * and corrupt what we are writing there. + */ + discard_lazy_cpu_state(); + +#ifdef CONFIG_ALTIVEC + regs->msr &= ~MSR_VEC; + if (msr & MSR_VEC) { + /* restore altivec registers from the stack */ + if (__copy_from_user(current->thread.vr, &sr->mc_vregs, + sizeof(sr->mc_vregs)) || + __copy_from_user(current->thread.transact_vr, + &tm_sr->mc_vregs, + sizeof(sr->mc_vregs))) + return 1; + } else if (current->thread.used_vr) { + memset(current->thread.vr, 0, ELF_NVRREG * sizeof(vector128)); + memset(current->thread.transact_vr, 0, + ELF_NVRREG * sizeof(vector128)); + } + + /* Always get VRSAVE back */ + if (__get_user(current->thread.vrsave, + (u32 __user *)&sr->mc_vregs[32]) || + __get_user(current->thread.transact_vrsave, + (u32 __user *)&tm_sr->mc_vregs[32])) + return 1; +#endif /* CONFIG_ALTIVEC */ + + regs->msr &= ~(MSR_FP | MSR_FE0 | MSR_FE1); + + if (copy_fpr_from_user(current, &sr->mc_fregs) || + copy_transact_fpr_from_user(current, &tm_sr->mc_fregs)) + return 1; + +#ifdef CONFIG_VSX + regs->msr &= ~MSR_VSX; + if (msr & MSR_VSX) { + /* + * Restore altivec registers from the stack to a local + * buffer, then write this out to the thread_struct + */ + if (copy_vsx_from_user(current, &sr->mc_vsregs) || + copy_transact_vsx_from_user(current, &tm_sr->mc_vsregs)) + return 1; + } else if (current->thread.used_vsr) + for (i = 0; i < 32 ; i++) { + current->thread.fpr[i][TS_VSRLOWOFFSET] = 0; + current->thread.transact_fpr[i][TS_VSRLOWOFFSET] = 0; + } +#endif /* CONFIG_VSX */ + +#ifdef CONFIG_SPE + /* SPE regs are not checkpointed with TM, so this section is + * simply the same as in restore_user_regs(). + */ + regs->msr &= ~MSR_SPE; + if (msr & MSR_SPE) { + if (__copy_from_user(current->thread.evr, &sr->mc_vregs, + ELF_NEVRREG * sizeof(u32))) + return 1; + } else if (current->thread.used_spe) + memset(current->thread.evr, 0, ELF_NEVRREG * sizeof(u32)); + + /* Always get SPEFSCR back */ + if (__get_user(current->thread.spefscr, (u32 __user *)&sr->mc_vregs + + ELF_NEVRREG)) + return 1; +#endif /* CONFIG_SPE */ + + /* Now, recheckpoint. This loads up all of the checkpointed (older) + * registers, including FP and V[S]Rs. After recheckpointing, the + * transactional versions should be loaded. + */ + tm_enable(); + /* This loads the checkpointed FP/VEC state, if used */ + tm_recheckpoint(¤t->thread, msr); + /* The task has moved into TM state S, so ensure MSR reflects this */ + regs->msr = (regs->msr & ~MSR_TS_MASK) | MSR_TS_S; + + /* This loads the speculative FP/VEC state, if used */ + if (msr & MSR_FP) { + do_load_up_transact_fpu(¤t->thread); + regs->msr |= (MSR_FP | current->thread.fpexc_mode); + } + if (msr & MSR_VEC) { + do_load_up_transact_altivec(¤t->thread); + regs->msr |= MSR_VEC; + } + + return 0; +} +#endif + #ifdef CONFIG_PPC64 long compat_sys_rt_sigaction(int sig, const struct sigaction32 __user *act, struct sigaction32 __user *oact, size_t sigsetsize) @@ -827,6 +1189,8 @@ int handle_rt_signal32(unsigned long sig, struct k_sigaction *ka, struct mcontext __user *frame; void __user *addr; unsigned long newsp = 0; + int sigret; + unsigned long tramp; /* Set up Signal Frame */ /* Put a Real Time Context onto stack */ @@ -838,7 +1202,6 @@ int handle_rt_signal32(unsigned long sig, struct k_sigaction *ka, /* Put the siginfo & fill in most of the ucontext */ if (copy_siginfo_to_user(&rt_sf->info, info) || __put_user(0, &rt_sf->uc.uc_flags) - || __put_user(0, &rt_sf->uc.uc_link) || __put_user(current->sas_ss_sp, &rt_sf->uc.uc_stack.ss_sp) || __put_user(sas_ss_flags(regs->gpr[1]), &rt_sf->uc.uc_stack.ss_flags) @@ -852,14 +1215,37 @@ int handle_rt_signal32(unsigned long sig, struct k_sigaction *ka, frame = &rt_sf->uc.uc_mcontext; addr = frame; if (vdso32_rt_sigtramp && current->mm->context.vdso_base) { - if (save_user_regs(regs, frame, 0, 1)) - goto badframe; - regs->link = current->mm->context.vdso_base + vdso32_rt_sigtramp; + sigret = 0; + tramp = current->mm->context.vdso_base + vdso32_rt_sigtramp; } else { - if (save_user_regs(regs, frame, __NR_rt_sigreturn, 1)) + sigret = __NR_rt_sigreturn; + tramp = (unsigned long) frame->tramp; + } + +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + if (MSR_TM_ACTIVE(regs->msr)) { + if (save_tm_user_regs(regs, &rt_sf->uc.uc_mcontext, + &rt_sf->uc_transact.uc_mcontext, sigret)) goto badframe; - regs->link = (unsigned long) frame->tramp; } + else +#endif + if (save_user_regs(regs, frame, sigret, 1)) + goto badframe; + regs->link = tramp; + +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + if (MSR_TM_ACTIVE(regs->msr)) { + if (__put_user((unsigned long)&rt_sf->uc_transact, + &rt_sf->uc.uc_link) + || __put_user(to_user_ptr(&rt_sf->uc_transact.uc_mcontext), + &rt_sf->uc_transact.uc_regs)) + goto badframe; + } + else +#endif + if (__put_user(0, &rt_sf->uc.uc_link)) + goto badframe; current->thread.fpscr.val = 0; /* turn off all fp exceptions */ @@ -878,6 +1264,13 @@ int handle_rt_signal32(unsigned long sig, struct k_sigaction *ka, regs->nip = (unsigned long) ka->sa.sa_handler; /* enter the signal handler in big-endian mode */ regs->msr &= ~MSR_LE; +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + /* Remove TM bits from thread's MSR. The MSR in the sigcontext + * just indicates to userland that we were doing a transaction, but we + * don't want to return in transactional state: + */ + regs->msr &= ~MSR_TS_MASK; +#endif return 1; badframe: @@ -925,6 +1318,35 @@ static int do_setcontext(struct ucontext __user *ucp, struct pt_regs *regs, int return 0; } +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +static int do_setcontext_tm(struct ucontext __user *ucp, + struct ucontext __user *tm_ucp, + struct pt_regs *regs) +{ + sigset_t set; + struct mcontext __user *mcp; + struct mcontext __user *tm_mcp; + u32 cmcp; + u32 tm_cmcp; + + if (get_sigset_t(&set, &ucp->uc_sigmask)) + return -EFAULT; + + if (__get_user(cmcp, &ucp->uc_regs) || + __get_user(tm_cmcp, &tm_ucp->uc_regs)) + return -EFAULT; + mcp = (struct mcontext __user *)(u64)cmcp; + tm_mcp = (struct mcontext __user *)(u64)tm_cmcp; + /* no need to check access_ok(mcp), since mcp < 4GB */ + + set_current_blocked(&set); + if (restore_tm_user_regs(regs, mcp, tm_mcp)) + return -EFAULT; + + return 0; +} +#endif + long sys_swapcontext(struct ucontext __user *old_ctx, struct ucontext __user *new_ctx, int ctx_size, int r6, int r7, int r8, struct pt_regs *regs) @@ -1020,7 +1442,12 @@ long sys_rt_sigreturn(int r3, int r4, int r5, int r6, int r7, int r8, struct pt_regs *regs) { struct rt_sigframe __user *rt_sf; - +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + struct ucontext __user *uc_transact; + unsigned long msr_hi; + unsigned long tmp; + int tm_restore = 0; +#endif /* Always make any pending restarted system calls return -EINTR */ current_thread_info()->restart_block.fn = do_no_restart_syscall; @@ -1028,6 +1455,34 @@ long sys_rt_sigreturn(int r3, int r4, int r5, int r6, int r7, int r8, (regs->gpr[1] + __SIGNAL_FRAMESIZE + 16); if (!access_ok(VERIFY_READ, rt_sf, sizeof(*rt_sf))) goto bad; +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + if (__get_user(tmp, &rt_sf->uc.uc_link)) + goto bad; + uc_transact = (struct ucontext __user *)(uintptr_t)tmp; + if (uc_transact) { + u32 cmcp; + struct mcontext __user *mcp; + + if (__get_user(cmcp, &uc_transact->uc_regs)) + return -EFAULT; + mcp = (struct mcontext __user *)(u64)cmcp; + /* The top 32 bits of the MSR are stashed in the transactional + * ucontext. */ + if (__get_user(msr_hi, &mcp->mc_gregs[PT_MSR])) + goto bad; + + if (MSR_TM_SUSPENDED(msr_hi<<32)) { + /* We only recheckpoint on return if we're + * transaction. + */ + tm_restore = 1; + if (do_setcontext_tm(&rt_sf->uc, uc_transact, regs)) + goto bad; + } + } + if (!tm_restore) + /* Fall through, for non-TM restore */ +#endif if (do_setcontext(&rt_sf->uc, regs, 1)) goto bad; @@ -1179,6 +1634,8 @@ int handle_signal32(unsigned long sig, struct k_sigaction *ka, struct sigcontext __user *sc; struct sigframe __user *frame; unsigned long newsp = 0; + int sigret; + unsigned long tramp; /* Set up Signal Frame */ frame = get_sigframe(ka, regs, sizeof(*frame), 1); @@ -1201,14 +1658,25 @@ int handle_signal32(unsigned long sig, struct k_sigaction *ka, goto badframe; if (vdso32_sigtramp && current->mm->context.vdso_base) { - if (save_user_regs(regs, &frame->mctx, 0, 1)) - goto badframe; - regs->link = current->mm->context.vdso_base + vdso32_sigtramp; + sigret = 0; + tramp = current->mm->context.vdso_base + vdso32_sigtramp; } else { - if (save_user_regs(regs, &frame->mctx, __NR_sigreturn, 1)) + sigret = __NR_sigreturn; + tramp = (unsigned long) frame->mctx.tramp; + } + +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + if (MSR_TM_ACTIVE(regs->msr)) { + if (save_tm_user_regs(regs, &frame->mctx, &frame->mctx_transact, + sigret)) goto badframe; - regs->link = (unsigned long) frame->mctx.tramp; } + else +#endif + if (save_user_regs(regs, &frame->mctx, sigret, 1)) + goto badframe; + + regs->link = tramp; current->thread.fpscr.val = 0; /* turn off all fp exceptions */ @@ -1223,7 +1691,13 @@ int handle_signal32(unsigned long sig, struct k_sigaction *ka, regs->nip = (unsigned long) ka->sa.sa_handler; /* enter the signal handler in big-endian mode */ regs->msr &= ~MSR_LE; - +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + /* Remove TM bits from thread's MSR. The MSR in the sigcontext + * just indicates to userland that we were doing a transaction, but we + * don't want to return in transactional state: + */ + regs->msr &= ~MSR_TS_MASK; +#endif return 1; badframe: diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c index 1ca045d44324..7a76ee48a952 100644 --- a/arch/powerpc/kernel/signal_64.c +++ b/arch/powerpc/kernel/signal_64.c @@ -34,6 +34,7 @@ #include <asm/syscalls.h> #include <asm/vdso.h> #include <asm/switch_to.h> +#include <asm/tm.h> #include "signal.h" @@ -56,6 +57,9 @@ struct rt_sigframe { /* sys_rt_sigreturn requires the ucontext be the first field */ struct ucontext uc; +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + struct ucontext uc_transact; +#endif unsigned long _unused[2]; unsigned int tramp[TRAMP_SIZE]; struct siginfo __user *pinfo; @@ -145,6 +149,145 @@ static long setup_sigcontext(struct sigcontext __user *sc, struct pt_regs *regs, return err; } +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +/* + * As above, but Transactional Memory is in use, so deliver sigcontexts + * containing checkpointed and transactional register states. + * + * To do this, we treclaim to gather both sets of registers and set up the + * 'normal' sigcontext registers with rolled-back register values such that a + * simple signal handler sees a correct checkpointed register state. + * If interested, a TM-aware sighandler can examine the transactional registers + * in the 2nd sigcontext to determine the real origin of the signal. + */ +static long setup_tm_sigcontexts(struct sigcontext __user *sc, + struct sigcontext __user *tm_sc, + struct pt_regs *regs, + int signr, sigset_t *set, unsigned long handler) +{ + /* When CONFIG_ALTIVEC is set, we _always_ setup v_regs even if the + * process never used altivec yet (MSR_VEC is zero in pt_regs of + * the context). This is very important because we must ensure we + * don't lose the VRSAVE content that may have been set prior to + * the process doing its first vector operation + * Userland shall check AT_HWCAP to know wether it can rely on the + * v_regs pointer or not. + */ +#ifdef CONFIG_ALTIVEC + elf_vrreg_t __user *v_regs = (elf_vrreg_t __user *) + (((unsigned long)sc->vmx_reserve + 15) & ~0xful); + elf_vrreg_t __user *tm_v_regs = (elf_vrreg_t __user *) + (((unsigned long)tm_sc->vmx_reserve + 15) & ~0xful); +#endif + unsigned long msr = regs->msr; + long err = 0; + + BUG_ON(!MSR_TM_ACTIVE(regs->msr)); + + /* tm_reclaim rolls back all reg states, saving checkpointed (older) + * GPRs to thread.ckpt_regs and (if used) FPRs to (newer) + * thread.transact_fp and/or VRs to (newer) thread.transact_vr. + * THEN we save out FP/VRs, if necessary, to the checkpointed (older) + * thread.fr[]/vr[]s. The transactional (newer) GPRs are on the + * stack, in *regs. + */ + tm_enable(); + tm_reclaim(¤t->thread, msr, TM_CAUSE_SIGNAL); + + flush_fp_to_thread(current); + +#ifdef CONFIG_ALTIVEC + err |= __put_user(v_regs, &sc->v_regs); + err |= __put_user(tm_v_regs, &tm_sc->v_regs); + + /* save altivec registers */ + if (current->thread.used_vr) { + flush_altivec_to_thread(current); + /* Copy 33 vec registers (vr0..31 and vscr) to the stack */ + err |= __copy_to_user(v_regs, current->thread.vr, + 33 * sizeof(vector128)); + /* If VEC was enabled there are transactional VRs valid too, + * else they're a copy of the checkpointed VRs. + */ + if (msr & MSR_VEC) + err |= __copy_to_user(tm_v_regs, + current->thread.transact_vr, + 33 * sizeof(vector128)); + else + err |= __copy_to_user(tm_v_regs, + current->thread.vr, + 33 * sizeof(vector128)); + + /* set MSR_VEC in the MSR value in the frame to indicate + * that sc->v_reg contains valid data. + */ + msr |= MSR_VEC; + } + /* We always copy to/from vrsave, it's 0 if we don't have or don't + * use altivec. + */ + err |= __put_user(current->thread.vrsave, (u32 __user *)&v_regs[33]); + if (msr & MSR_VEC) + err |= __put_user(current->thread.transact_vrsave, + (u32 __user *)&tm_v_regs[33]); + else + err |= __put_user(current->thread.vrsave, + (u32 __user *)&tm_v_regs[33]); + +#else /* CONFIG_ALTIVEC */ + err |= __put_user(0, &sc->v_regs); + err |= __put_user(0, &tm_sc->v_regs); +#endif /* CONFIG_ALTIVEC */ + + /* copy fpr regs and fpscr */ + err |= copy_fpr_to_user(&sc->fp_regs, current); + if (msr & MSR_FP) + err |= copy_transact_fpr_to_user(&tm_sc->fp_regs, current); + else + err |= copy_fpr_to_user(&tm_sc->fp_regs, current); + +#ifdef CONFIG_VSX + /* + * Copy VSX low doubleword to local buffer for formatting, + * then out to userspace. Update v_regs to point after the + * VMX data. + */ + if (current->thread.used_vsr) { + __giveup_vsx(current); + v_regs += ELF_NVRREG; + tm_v_regs += ELF_NVRREG; + + err |= copy_vsx_to_user(v_regs, current); + + if (msr & MSR_VSX) + err |= copy_transact_vsx_to_user(tm_v_regs, current); + else + err |= copy_vsx_to_user(tm_v_regs, current); + + /* set MSR_VSX in the MSR value in the frame to + * indicate that sc->vs_reg) contains valid data. + */ + msr |= MSR_VSX; + } +#endif /* CONFIG_VSX */ + + err |= __put_user(&sc->gp_regs, &sc->regs); + err |= __put_user(&tm_sc->gp_regs, &tm_sc->regs); + WARN_ON(!FULL_REGS(regs)); + err |= __copy_to_user(&tm_sc->gp_regs, regs, GP_REGS_SIZE); + err |= __copy_to_user(&sc->gp_regs, + ¤t->thread.ckpt_regs, GP_REGS_SIZE); + err |= __put_user(msr, &tm_sc->gp_regs[PT_MSR]); + err |= __put_user(msr, &sc->gp_regs[PT_MSR]); + err |= __put_user(signr, &sc->signal); + err |= __put_user(handler, &sc->handler); + if (set != NULL) + err |= __put_user(set->sig[0], &sc->oldmask); + + return err; +} +#endif + /* * Restore the sigcontext from the signal frame. */ @@ -241,6 +384,153 @@ static long restore_sigcontext(struct pt_regs *regs, sigset_t *set, int sig, return err; } +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM +/* + * Restore the two sigcontexts from the frame of a transactional processes. + */ + +static long restore_tm_sigcontexts(struct pt_regs *regs, + struct sigcontext __user *sc, + struct sigcontext __user *tm_sc) +{ +#ifdef CONFIG_ALTIVEC + elf_vrreg_t __user *v_regs, *tm_v_regs; +#endif + unsigned long err = 0; + unsigned long msr; +#ifdef CONFIG_VSX + int i; +#endif + /* copy the GPRs */ + err |= __copy_from_user(regs->gpr, tm_sc->gp_regs, sizeof(regs->gpr)); + err |= __copy_from_user(¤t->thread.ckpt_regs, sc->gp_regs, + sizeof(regs->gpr)); + + /* + * TFHAR is restored from the checkpointed 'wound-back' ucontext's NIP. + * TEXASR was set by the signal delivery reclaim, as was TFIAR. + * Users doing anything abhorrent like thread-switching w/ signals for + * TM-Suspended code will have to back TEXASR/TFIAR up themselves. + * For the case of getting a signal and simply returning from it, + * we don't need to re-copy them here. + */ + err |= __get_user(regs->nip, &tm_sc->gp_regs[PT_NIP]); + err |= __get_user(current->thread.tm_tfhar, &sc->gp_regs[PT_NIP]); + + /* get MSR separately, transfer the LE bit if doing signal return */ + err |= __get_user(msr, &sc->gp_regs[PT_MSR]); + regs->msr = (regs->msr & ~MSR_LE) | (msr & MSR_LE); + + /* The following non-GPR non-FPR non-VR state is also checkpointed: */ + err |= __get_user(regs->ctr, &tm_sc->gp_regs[PT_CTR]); + err |= __get_user(regs->link, &tm_sc->gp_regs[PT_LNK]); + err |= __get_user(regs->xer, &tm_sc->gp_regs[PT_XER]); + err |= __get_user(regs->ccr, &tm_sc->gp_regs[PT_CCR]); + err |= __get_user(current->thread.ckpt_regs.ctr, + &sc->gp_regs[PT_CTR]); + err |= __get_user(current->thread.ckpt_regs.link, + &sc->gp_regs[PT_LNK]); + err |= __get_user(current->thread.ckpt_regs.xer, + &sc->gp_regs[PT_XER]); + err |= __get_user(current->thread.ckpt_regs.ccr, + &sc->gp_regs[PT_CCR]); + + /* These regs are not checkpointed; they can go in 'regs'. */ + err |= __get_user(regs->trap, &sc->gp_regs[PT_TRAP]); + err |= __get_user(regs->dar, &sc->gp_regs[PT_DAR]); + err |= __get_user(regs->dsisr, &sc->gp_regs[PT_DSISR]); + err |= __get_user(regs->result, &sc->gp_regs[PT_RESULT]); + + /* + * Do this before updating the thread state in + * current->thread.fpr/vr. That way, if we get preempted + * and another task grabs the FPU/Altivec, it won't be + * tempted to save the current CPU state into the thread_struct + * and corrupt what we are writing there. + */ + discard_lazy_cpu_state(); + + /* + * Force reload of FP/VEC. + * This has to be done before copying stuff into current->thread.fpr/vr + * for the reasons explained in the previous comment. + */ + regs->msr &= ~(MSR_FP | MSR_FE0 | MSR_FE1 | MSR_VEC | MSR_VSX); + +#ifdef CONFIG_ALTIVEC + err |= __get_user(v_regs, &sc->v_regs); + err |= __get_user(tm_v_regs, &tm_sc->v_regs); + if (err) + return err; + if (v_regs && !access_ok(VERIFY_READ, v_regs, 34 * sizeof(vector128))) + return -EFAULT; + if (tm_v_regs && !access_ok(VERIFY_READ, + tm_v_regs, 34 * sizeof(vector128))) + return -EFAULT; + /* Copy 33 vec registers (vr0..31 and vscr) from the stack */ + if (v_regs != 0 && tm_v_regs != 0 && (msr & MSR_VEC) != 0) { + err |= __copy_from_user(current->thread.vr, v_regs, + 33 * sizeof(vector128)); + err |= __copy_from_user(current->thread.transact_vr, tm_v_regs, + 33 * sizeof(vector128)); + } + else if (current->thread.used_vr) { + memset(current->thread.vr, 0, 33 * sizeof(vector128)); + memset(current->thread.transact_vr, 0, 33 * sizeof(vector128)); + } + /* Always get VRSAVE back */ + if (v_regs != 0 && tm_v_regs != 0) { + err |= __get_user(current->thread.vrsave, + (u32 __user *)&v_regs[33]); + err |= __get_user(current->thread.transact_vrsave, + (u32 __user *)&tm_v_regs[33]); + } + else { + current->thread.vrsave = 0; + current->thread.transact_vrsave = 0; + } +#endif /* CONFIG_ALTIVEC */ + /* restore floating point */ + err |= copy_fpr_from_user(current, &sc->fp_regs); + err |= copy_transact_fpr_from_user(current, &tm_sc->fp_regs); +#ifdef CONFIG_VSX + /* + * Get additional VSX data. Update v_regs to point after the + * VMX data. Copy VSX low doubleword from userspace to local + * buffer for formatting, then into the taskstruct. + */ + if (v_regs && ((msr & MSR_VSX) != 0)) { + v_regs += ELF_NVRREG; + tm_v_regs += ELF_NVRREG; + err |= copy_vsx_from_user(current, v_regs); + err |= copy_transact_vsx_from_user(current, tm_v_regs); + } else { + for (i = 0; i < 32 ; i++) { + current->thread.fpr[i][TS_VSRLOWOFFSET] = 0; + current->thread.transact_fpr[i][TS_VSRLOWOFFSET] = 0; + } + } +#endif + tm_enable(); + /* This loads the checkpointed FP/VEC state, if used */ + tm_recheckpoint(¤t->thread, msr); + /* The task has moved into TM state S, so ensure MSR reflects this: */ + regs->msr = (regs->msr & ~MSR_TS_MASK) | __MASK(33); + + /* This loads the speculative FP/VEC state, if used */ + if (msr & MSR_FP) { + do_load_up_transact_fpu(¤t->thread); + regs->msr |= (MSR_FP | current->thread.fpexc_mode); + } + if (msr & MSR_VEC) { + do_load_up_transact_altivec(¤t->thread); + regs->msr |= MSR_VEC; + } + + return err; +} +#endif + /* * Setup the trampoline code on the stack */ @@ -355,6 +645,9 @@ int sys_rt_sigreturn(unsigned long r3, unsigned long r4, unsigned long r5, { struct ucontext __user *uc = (struct ucontext __user *)regs->gpr[1]; sigset_t set; +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + unsigned long msr; +#endif /* Always make any pending restarted system calls return -EINTR */ current_thread_info()->restart_block.fn = do_no_restart_syscall; @@ -365,6 +658,21 @@ int sys_rt_sigreturn(unsigned long r3, unsigned long r4, unsigned long r5, if (__copy_from_user(&set, &uc->uc_sigmask, sizeof(set))) goto badframe; set_current_blocked(&set); +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM + if (__get_user(msr, &uc->uc_mcontext.gp_regs[PT_MSR])) + goto badframe; + if (MSR_TM_SUSPENDED(msr)) { + /* We recheckpoint on return. */ + struct ucontext __user *uc_transact; + if (__get_user(uc_transact, &uc->uc_link)) + goto badframe; + if (restore_tm_sigcontexts(regs, &uc->uc_ |