* [PATCH 03/17] powerpc: Add additional state needed for transactional memory to thread struct
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Set of new archtected state for saving away on context switch.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/include/asm/processor.h | 28 ++++++++++++++++++++++++++++
1 file changed, 28 insertions(+)
diff --git a/arch/powerpc/include/asm/processor.h b/arch/powerpc/include/asm/processor.h
index 7938658..1ffed17 100644
--- a/arch/powerpc/include/asm/processor.h
+++ b/arch/powerpc/include/asm/processor.h
@@ -246,6 +246,34 @@ struct thread_struct {
unsigned long spefscr; /* SPE & eFP status */
int used_spe; /* set if process has used spe */
#endif /* CONFIG_SPE */
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ u64 tm_tfhar; /* Transaction fail handler addr */
+ u64 tm_texasr; /* Transaction exception & summary */
+ u64 tm_tfiar; /* Transaction fail instr address reg */
+ unsigned long tm_orig_msr; /* Thread's MSR on ctx switch */
+ struct pt_regs ckpt_regs; /* Checkpointed registers */
+
+ /*
+ * Transactional FP and VSX 0-31 register set.
+ * NOTE: the sense of these is the opposite of the integer ckpt_regs!
+ *
+ * When a transaction is active/signalled/scheduled etc., *regs is the
+ * most recent set of/speculated GPRs with ckpt_regs being the older
+ * checkpointed regs to which we roll back if transaction aborts.
+ *
+ * However, fpr[] is the checkpointed 'base state' of FP regs, and
+ * transact_fpr[] is the new set of transactional values.
+ * VRs work the same way.
+ */
+ double transact_fpr[32][TS_FPRWIDTH];
+ struct {
+ unsigned int pad;
+ unsigned int val; /* Floating point status */
+ } transact_fpscr;
+ vector128 transact_vr[32] __attribute__((aligned(16)));
+ vector128 transact_vscr __attribute__((aligned(16)));
+ unsigned long transact_vrsave;
+#endif /* CONFIG_TRANSACTIONAL_MEM */
#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
void* kvm_shadow_vcpu; /* KVM internal data */
#endif /* CONFIG_KVM_BOOK3S_32_HANDLER */
--
1.7.10.4
^ permalink raw reply related
* [PATCH 04/17] powerpc: New macros for transactional memory support
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
This adds new macros for saving and restoring checkpointed architected state
from and to the thread_struct.
It also adds some debugging macros for when your brain explodes trying to debug
your transactional memory enabled kernel.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/include/asm/ppc_asm.h | 83 ++++++++++++++++++++++++++++++++++
arch/powerpc/include/asm/processor.h | 1 +
arch/powerpc/kernel/asm-offsets.c | 24 ++++++++++
arch/powerpc/kernel/fpu.S | 12 +++++
arch/powerpc/kernel/process.c | 10 ++++
arch/powerpc/kernel/traps.c | 11 +++++
6 files changed, 141 insertions(+)
diff --git a/arch/powerpc/include/asm/ppc_asm.h b/arch/powerpc/include/asm/ppc_asm.h
index c2d0e58..54219ce 100644
--- a/arch/powerpc/include/asm/ppc_asm.h
+++ b/arch/powerpc/include/asm/ppc_asm.h
@@ -123,6 +123,89 @@ END_FW_FTR_SECTION_IFSET(FW_FEATURE_SPLPAR)
#define REST_16VRS(n,b,base) REST_8VRS(n,b,base); REST_8VRS(n+8,b,base)
#define REST_32VRS(n,b,base) REST_16VRS(n,b,base); REST_16VRS(n+16,b,base)
+/* Save/restore FPRs, VRs and VSRs from their checkpointed backups in
+ * thread_struct:
+ */
+#define SAVE_FPR_TRANSACT(n, base) stfd n,THREAD_TRANSACT_FPR0+ \
+ 8*TS_FPRWIDTH*(n)(base)
+#define SAVE_2FPRS_TRANSACT(n, base) SAVE_FPR_TRANSACT(n, base); \
+ SAVE_FPR_TRANSACT(n+1, base)
+#define SAVE_4FPRS_TRANSACT(n, base) SAVE_2FPRS_TRANSACT(n, base); \
+ SAVE_2FPRS_TRANSACT(n+2, base)
+#define SAVE_8FPRS_TRANSACT(n, base) SAVE_4FPRS_TRANSACT(n, base); \
+ SAVE_4FPRS_TRANSACT(n+4, base)
+#define SAVE_16FPRS_TRANSACT(n, base) SAVE_8FPRS_TRANSACT(n, base); \
+ SAVE_8FPRS_TRANSACT(n+8, base)
+#define SAVE_32FPRS_TRANSACT(n, base) SAVE_16FPRS_TRANSACT(n, base); \
+ SAVE_16FPRS_TRANSACT(n+16, base)
+
+#define REST_FPR_TRANSACT(n, base) lfd n,THREAD_TRANSACT_FPR0+ \
+ 8*TS_FPRWIDTH*(n)(base)
+#define REST_2FPRS_TRANSACT(n, base) REST_FPR_TRANSACT(n, base); \
+ REST_FPR_TRANSACT(n+1, base)
+#define REST_4FPRS_TRANSACT(n, base) REST_2FPRS_TRANSACT(n, base); \
+ REST_2FPRS_TRANSACT(n+2, base)
+#define REST_8FPRS_TRANSACT(n, base) REST_4FPRS_TRANSACT(n, base); \
+ REST_4FPRS_TRANSACT(n+4, base)
+#define REST_16FPRS_TRANSACT(n, base) REST_8FPRS_TRANSACT(n, base); \
+ REST_8FPRS_TRANSACT(n+8, base)
+#define REST_32FPRS_TRANSACT(n, base) REST_16FPRS_TRANSACT(n, base); \
+ REST_16FPRS_TRANSACT(n+16, base)
+
+
+#define SAVE_VR_TRANSACT(n,b,base) li b,THREAD_TRANSACT_VR0+(16*(n)); \
+ stvx n,b,base
+#define SAVE_2VRS_TRANSACT(n,b,base) SAVE_VR_TRANSACT(n,b,base); \
+ SAVE_VR_TRANSACT(n+1,b,base)
+#define SAVE_4VRS_TRANSACT(n,b,base) SAVE_2VRS_TRANSACT(n,b,base); \
+ SAVE_2VRS_TRANSACT(n+2,b,base)
+#define SAVE_8VRS_TRANSACT(n,b,base) SAVE_4VRS_TRANSACT(n,b,base); \
+ SAVE_4VRS_TRANSACT(n+4,b,base)
+#define SAVE_16VRS_TRANSACT(n,b,base) SAVE_8VRS_TRANSACT(n,b,base); \
+ SAVE_8VRS_TRANSACT(n+8,b,base)
+#define SAVE_32VRS_TRANSACT(n,b,base) SAVE_16VRS_TRANSACT(n,b,base); \
+ SAVE_16VRS_TRANSACT(n+16,b,base)
+
+#define REST_VR_TRANSACT(n,b,base) li b,THREAD_TRANSACT_VR0+(16*(n)); \
+ lvx n,b,base
+#define REST_2VRS_TRANSACT(n,b,base) REST_VR_TRANSACT(n,b,base); \
+ REST_VR_TRANSACT(n+1,b,base)
+#define REST_4VRS_TRANSACT(n,b,base) REST_2VRS_TRANSACT(n,b,base); \
+ REST_2VRS_TRANSACT(n+2,b,base)
+#define REST_8VRS_TRANSACT(n,b,base) REST_4VRS_TRANSACT(n,b,base); \
+ REST_4VRS_TRANSACT(n+4,b,base)
+#define REST_16VRS_TRANSACT(n,b,base) REST_8VRS_TRANSACT(n,b,base); \
+ REST_8VRS_TRANSACT(n+8,b,base)
+#define REST_32VRS_TRANSACT(n,b,base) REST_16VRS_TRANSACT(n,b,base); \
+ REST_16VRS_TRANSACT(n+16,b,base)
+
+
+#define SAVE_VSR_TRANSACT(n,b,base) li b,THREAD_TRANSACT_VSR0+(16*(n)); \
+ STXVD2X(n,R##base,R##b)
+#define SAVE_2VSRS_TRANSACT(n,b,base) SAVE_VSR_TRANSACT(n,b,base); \
+ SAVE_VSR_TRANSACT(n+1,b,base)
+#define SAVE_4VSRS_TRANSACT(n,b,base) SAVE_2VSRS_TRANSACT(n,b,base); \
+ SAVE_2VSRS_TRANSACT(n+2,b,base)
+#define SAVE_8VSRS_TRANSACT(n,b,base) SAVE_4VSRS_TRANSACT(n,b,base); \
+ SAVE_4VSRS_TRANSACT(n+4,b,base)
+#define SAVE_16VSRS_TRANSACT(n,b,base) SAVE_8VSRS_TRANSACT(n,b,base); \
+ SAVE_8VSRS_TRANSACT(n+8,b,base)
+#define SAVE_32VSRS_TRANSACT(n,b,base) SAVE_16VSRS_TRANSACT(n,b,base); \
+ SAVE_16VSRS_TRANSACT(n+16,b,base)
+
+#define REST_VSR_TRANSACT(n,b,base) li b,THREAD_TRANSACT_VSR0+(16*(n)); \
+ LXVD2X(n,R##base,R##b)
+#define REST_2VSRS_TRANSACT(n,b,base) REST_VSR_TRANSACT(n,b,base); \
+ REST_VSR_TRANSACT(n+1,b,base)
+#define REST_4VSRS_TRANSACT(n,b,base) REST_2VSRS_TRANSACT(n,b,base); \
+ REST_2VSRS_TRANSACT(n+2,b,base)
+#define REST_8VSRS_TRANSACT(n,b,base) REST_4VSRS_TRANSACT(n,b,base); \
+ REST_4VSRS_TRANSACT(n+4,b,base)
+#define REST_16VSRS_TRANSACT(n,b,base) REST_8VSRS_TRANSACT(n,b,base); \
+ REST_8VSRS_TRANSACT(n+8,b,base)
+#define REST_32VSRS_TRANSACT(n,b,base) REST_16VSRS_TRANSACT(n,b,base); \
+ REST_16VSRS_TRANSACT(n+16,b,base)
+
/* Save the lower 32 VSRs in the thread VSR region */
#define SAVE_VSR(n,b,base) li b,THREAD_VSR0+(16*(n)); STXVD2X(n,R##base,R##b)
#define SAVE_2VSRS(n,b,base) SAVE_VSR(n,b,base); SAVE_VSR(n+1,b,base)
diff --git a/arch/powerpc/include/asm/processor.h b/arch/powerpc/include/asm/processor.h
index 1ffed17..af420e8 100644
--- a/arch/powerpc/include/asm/processor.h
+++ b/arch/powerpc/include/asm/processor.h
@@ -152,6 +152,7 @@ typedef struct {
#define TS_FPROFFSET 0
#define TS_VSRLOWOFFSET 1
#define TS_FPR(i) fpr[i][TS_FPROFFSET]
+#define TS_TRANS_FPR(i) transact_fpr[i][TS_FPROFFSET]
struct thread_struct {
unsigned long ksp; /* Kernel stack pointer */
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index e39ca55..3b81960 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -122,6 +122,30 @@ int main(void)
DEFINE(THREAD_KVM_VCPU, offsetof(struct thread_struct, kvm_vcpu));
#endif
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ DEFINE(THREAD_TM_TFHAR, offsetof(struct thread_struct, tm_tfhar));
+ DEFINE(THREAD_TM_TEXASR, offsetof(struct thread_struct, tm_texasr));
+ DEFINE(THREAD_TM_TFIAR, offsetof(struct thread_struct, tm_tfiar));
+ DEFINE(PT_CKPT_REGS, offsetof(struct thread_struct, ckpt_regs));
+ DEFINE(THREAD_TRANSACT_VR0, offsetof(struct thread_struct,
+ transact_vr[0]));
+ DEFINE(THREAD_TRANSACT_VSCR, offsetof(struct thread_struct,
+ transact_vscr));
+ DEFINE(THREAD_TRANSACT_VRSAVE, offsetof(struct thread_struct,
+ transact_vrsave));
+ DEFINE(THREAD_TRANSACT_FPR0, offsetof(struct thread_struct,
+ transact_fpr[0]));
+ DEFINE(THREAD_TRANSACT_FPSCR, offsetof(struct thread_struct,
+ transact_fpscr));
+#ifdef CONFIG_VSX
+ DEFINE(THREAD_TRANSACT_VSR0, offsetof(struct thread_struct,
+ transact_fpr[0]));
+#endif
+ /* Local pt_regs on stack for Transactional Memory funcs. */
+ DEFINE(TM_FRAME_SIZE, STACK_FRAME_OVERHEAD +
+ sizeof(struct pt_regs) + 16);
+#endif /* CONFIG_TRANSACTIONAL_MEM */
+
DEFINE(TI_FLAGS, offsetof(struct thread_info, flags));
DEFINE(TI_LOCAL_FLAGS, offsetof(struct thread_info, local_flags));
DEFINE(TI_PREEMPT, offsetof(struct thread_info, preempt_count));
diff --git a/arch/powerpc/kernel/fpu.S b/arch/powerpc/kernel/fpu.S
index e0ada05..adb1551 100644
--- a/arch/powerpc/kernel/fpu.S
+++ b/arch/powerpc/kernel/fpu.S
@@ -35,6 +35,15 @@ END_FTR_SECTION_IFSET(CPU_FTR_VSX); \
2: REST_32VSRS(n,c,base); \
3:
+#define __REST_32FPVSRS_TRANSACT(n,c,base) \
+BEGIN_FTR_SECTION \
+ b 2f; \
+END_FTR_SECTION_IFSET(CPU_FTR_VSX); \
+ REST_32FPRS_TRANSACT(n,base); \
+ b 3f; \
+2: REST_32VSRS_TRANSACT(n,c,base); \
+3:
+
#define __SAVE_32FPVSRS(n,c,base) \
BEGIN_FTR_SECTION \
b 2f; \
@@ -45,9 +54,12 @@ END_FTR_SECTION_IFSET(CPU_FTR_VSX); \
3:
#else
#define __REST_32FPVSRS(n,b,base) REST_32FPRS(n, base)
+#define __REST_32FPVSRS_TRANSACT(n,b,base) REST_32FPRS(n, base)
#define __SAVE_32FPVSRS(n,b,base) SAVE_32FPRS(n, base)
#endif
#define REST_32FPVSRS(n,c,base) __REST_32FPVSRS(n,__REG_##c,__REG_##base)
+#define REST_32FPVSRS_TRANSACT(n,c,base) \
+ __REST_32FPVSRS_TRANSACT(n,__REG_##c,__REG_##base)
#define SAVE_32FPVSRS(n,c,base) __SAVE_32FPVSRS(n,__REG_##c,__REG_##base)
/*
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index 99550d3..1ab3a2c 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -57,6 +57,16 @@
#include <linux/kprobes.h>
#include <linux/kdebug.h>
+#ifdef CONFIG_TRANSACTIONAL_MEM
+/* Transactional Memory debug */
+// #define TM_DEBUG_SW
+#ifdef TM_DEBUG_SW
+#define TM_DEBUG(x...) printk(KERN_INFO x)
+#else
+#define TM_DEBUG(x...) do { } while(0)
+#endif
+#endif
+
extern unsigned long _get_SP(void);
#ifndef CONFIG_SMP
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index a008cf5..2a62f54 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -78,6 +78,16 @@ EXPORT_SYMBOL(__debugger_break_match);
EXPORT_SYMBOL(__debugger_fault_handler);
#endif
+#ifdef CONFIG_TRANSACTIONAL_MEM
+/* Transactional Memory trap debug */
+// #define TM_DEBUG_SW
+#ifdef TM_DEBUG_SW
+#define TM_DEBUG(x...) printk(KERN_INFO x)
+#else
+#define TM_DEBUG(x...) do { } while(0)
+#endif
+#endif
+
/*
* Trap & Exception support
*/
@@ -350,6 +360,7 @@ static inline int check_io_access(struct pt_regs *regs)
exception is in the MSR. */
#define get_reason(regs) ((regs)->msr)
#define get_mc_reason(regs) ((regs)->msr)
+#define REASON_TM 0x200000
#define REASON_FP 0x100000
#define REASON_ILLEGAL 0x80000
#define REASON_PRIVILEGED 0x40000
--
1.7.10.4
^ permalink raw reply related
* [PATCH 05/17] powerpc: Register defines for various transactional memory registers
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Defines for MSR bits and transactional memory related SPRs TFIAR, TEXASR and
TEXASRU.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/include/asm/reg.h | 27 +++++++++++++++++++++++++++
1 file changed, 27 insertions(+)
diff --git a/arch/powerpc/include/asm/reg.h b/arch/powerpc/include/asm/reg.h
index 1f59fbb..3acef78 100644
--- a/arch/powerpc/include/asm/reg.h
+++ b/arch/powerpc/include/asm/reg.h
@@ -29,6 +29,10 @@
#define MSR_SF_LG 63 /* Enable 64 bit mode */
#define MSR_ISF_LG 61 /* Interrupt 64b mode valid on 630 */
#define MSR_HV_LG 60 /* Hypervisor state */
+#define MSR_TS_T_LG 34 /* Trans Mem state: Transactional */
+#define MSR_TS_S_LG 33 /* Trans Mem state: Suspended */
+#define MSR_TS_LG 33 /* Trans Mem state (2 bits) */
+#define MSR_TM_LG 32 /* Trans Mem Available */
#define MSR_VEC_LG 25 /* Enable AltiVec */
#define MSR_VSX_LG 23 /* Enable VSX */
#define MSR_POW_LG 18 /* Enable Power Management */
@@ -98,6 +102,25 @@
#define MSR_RI __MASK(MSR_RI_LG) /* Recoverable Exception */
#define MSR_LE __MASK(MSR_LE_LG) /* Little Endian */
+#define MSR_TM __MASK(MSR_TM_LG) /* Transactional Mem Available */
+#define MSR_TS_N 0 /* Non-transactional */
+#define MSR_TS_S __MASK(MSR_TS_S_LG) /* Transaction Suspended */
+#define MSR_TS_T __MASK(MSR_TS_T_LG) /* Transaction Transactional */
+#define MSR_TS_MASK (MSR_TS_T | MSR_TS_S) /* Transaction State bits */
+#define MSR_TM_ACTIVE(x) (((x) & MSR_TS_MASK) != 0) /* Transaction active? */
+#define MSR_TM_TRANSACTIONAL(x) (((x) & MSR_TS_MASK) == MSR_TS_T)
+#define MSR_TM_SUSPENDED(x) (((x) & MSR_TS_MASK) == MSR_TS_S)
+
+/* Reason codes describing kernel causes for transaction aborts. By
+ convention, bit0 is copied to TEXASR[56] (IBM bit 7) which is set if
+ the failure is persistent.
+*/
+#define TM_CAUSE_RESCHED 0xfe
+#define TM_CAUSE_TLBI 0xfc
+#define TM_CAUSE_FAC_UNAV 0xfa
+#define TM_CAUSE_SYSCALL 0xf9 /* Persistent */
+#define TM_CAUSE_MISC 0xf6
+
#if defined(CONFIG_PPC_BOOK3S_64)
#define MSR_64BIT MSR_SF
@@ -193,6 +216,10 @@
#define SPRN_UAMOR 0x9d /* User Authority Mask Override Register */
#define SPRN_AMOR 0x15d /* Authority Mask Override Register */
#define SPRN_ACOP 0x1F /* Available Coprocessor Register */
+#define SPRN_TFIAR 0x81 /* Transaction Failure Inst Addr */
+#define SPRN_TEXASR 0x82 /* Transaction EXception & Summary */
+#define SPRN_TEXASRU 0x83 /* '' '' '' Upper 32 */
+#define SPRN_TFHAR 0x80 /* Transaction Failure Handler Addr */
#define SPRN_CTRLF 0x088
#define SPRN_CTRLT 0x098
#define CTRL_CT 0xc0000000 /* current thread */
--
1.7.10.4
^ permalink raw reply related
* [PATCH 06/17] powerpc: Add transactional memory paca scratch register to show_regs
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Add transactional memory paca scratch register to show_regs. This is useful
for debugging.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/include/asm/paca.h | 1 +
arch/powerpc/kernel/asm-offsets.c | 1 +
arch/powerpc/kernel/entry_64.S | 4 ++++
arch/powerpc/kernel/process.c | 3 +++
4 files changed, 9 insertions(+)
diff --git a/arch/powerpc/include/asm/paca.h b/arch/powerpc/include/asm/paca.h
index c47d687..07e9851 100644
--- a/arch/powerpc/include/asm/paca.h
+++ b/arch/powerpc/include/asm/paca.h
@@ -137,6 +137,7 @@ struct paca_struct {
u8 irq_work_pending; /* IRQ_WORK interrupt while soft-disable */
u8 nap_state_lost; /* NV GPR values lost in power7_idle */
u64 sprg3; /* Saved user-visible sprg */
+ u64 tm_scratch; /* TM scratch area for reclaim */
#ifdef CONFIG_PPC_POWERNV
/* Pointer to OPAL machine check event structure set by the
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 3b81960..c9cdec4 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -123,6 +123,7 @@ int main(void)
#endif
#ifdef CONFIG_TRANSACTIONAL_MEM
+ DEFINE(PACATMSCRATCH, offsetof(struct paca_struct, tm_scratch));
DEFINE(THREAD_TM_TFHAR, offsetof(struct thread_struct, tm_tfhar));
DEFINE(THREAD_TM_TEXASR, offsetof(struct thread_struct, tm_texasr));
DEFINE(THREAD_TM_TFIAR, offsetof(struct thread_struct, tm_tfiar));
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index 6faf51e..283bc61 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -752,6 +752,10 @@ fast_exception_return:
andc r4,r4,r0 /* r0 contains MSR_RI here */
mtmsrd r4,1
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ /* TM debug */
+ std r3, PACATMSCRATCH(r13) /* Stash returned-to MSR */
+#endif
/*
* r13 is our per cpu area, only restore it if we are returning to
* userspace the value stored in the stack frame may belong to
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index 1ab3a2c..cb11fed 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -749,6 +749,9 @@ void show_regs(struct pt_regs * regs)
printk("NIP ["REG"] %pS\n", regs->nip, (void *)regs->nip);
printk("LR ["REG"] %pS\n", regs->link, (void *)regs->link);
#endif
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ printk("PACATMSCRATCH [%llx]\n", get_paca()->tm_scratch);
+#endif
show_stack(current, (unsigned long *) regs->gpr[1]);
if (!user_mode(regs))
show_instructions(regs);
--
1.7.10.4
^ permalink raw reply related
* [PATCH 07/17] powerpc: Add helper functions for transactional memory context switching
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Here we add the helper functions to be used when context switching. These
allow us to fully reclaim and recheckpoint a transaction.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/include/asm/reg.h | 2 +-
arch/powerpc/include/asm/tm.h | 20 +++
arch/powerpc/kernel/Makefile | 2 +
arch/powerpc/kernel/tm.S | 383 ++++++++++++++++++++++++++++++++++++++++
4 files changed, 406 insertions(+), 1 deletion(-)
create mode 100644 arch/powerpc/include/asm/tm.h
create mode 100644 arch/powerpc/kernel/tm.S
diff --git a/arch/powerpc/include/asm/reg.h b/arch/powerpc/include/asm/reg.h
index 3acef78..d19581ae4 100644
--- a/arch/powerpc/include/asm/reg.h
+++ b/arch/powerpc/include/asm/reg.h
@@ -794,7 +794,7 @@
* HV mode in which case it is HSPRG0
*
* 64-bit server:
- * - SPRG0 unused (reserved for HV on Power4)
+ * - SPRG0 scratch for TM recheckpoint/reclaim (reserved for HV on Power4)
* - SPRG2 scratch for exception vectors
* - SPRG3 CPU and NUMA node for VDSO getcpu (user visible)
* - HSPRG0 stores PACA in HV mode
diff --git a/arch/powerpc/include/asm/tm.h b/arch/powerpc/include/asm/tm.h
new file mode 100644
index 0000000..e9ad010
--- /dev/null
+++ b/arch/powerpc/include/asm/tm.h
@@ -0,0 +1,20 @@
+/*
+ * Transactional memory support routines to reclaim and recheckpoint
+ * transactional process state.
+ *
+ * Copyright 2012 Matt Evans & Michael Neuling, IBM Corporation.
+ */
+
+#ifdef CONFIG_TRANSACTIONAL_MEM
+extern void do_load_up_transact_fpu(struct thread_struct *thread);
+extern void do_load_up_transact_altivec(struct thread_struct *thread);
+#endif
+
+extern void tm_enable(void);
+extern void tm_reclaim(struct thread_struct *thread,
+ unsigned long orig_msr, uint8_t cause);
+extern void tm_recheckpoint(struct thread_struct *thread,
+ unsigned long orig_msr);
+extern void tm_abort(uint8_t cause);
+extern void tm_save_sprs(struct thread_struct *thread);
+extern void tm_restore_sprs(struct thread_struct *thread);
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index b4f0e36..7554968 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -121,6 +121,8 @@ ifneq ($(CONFIG_PPC_INDIRECT_IO),y)
obj-y += iomap.o
endif
+obj64-$(CONFIG_TRANSACTIONAL_MEM) += tm.o
+
obj-$(CONFIG_PPC64) += $(obj64-y)
obj-$(CONFIG_PPC32) += $(obj32-y)
diff --git a/arch/powerpc/kernel/tm.S b/arch/powerpc/kernel/tm.S
new file mode 100644
index 0000000..0517b08
--- /dev/null
+++ b/arch/powerpc/kernel/tm.S
@@ -0,0 +1,383 @@
+/*
+ * Transactional memory support routines to reclaim and recheckpoint
+ * transactional process state.
+ *
+ * Copyright 2012 Matt Evans & Michael Neuling, IBM Corporation.
+ */
+
+#include <asm/asm-offsets.h>
+#include <asm/ppc_asm.h>
+#include <asm/ppc-opcode.h>
+#include <asm/ptrace.h>
+#include <asm/reg.h>
+
+#ifdef CONFIG_VSX
+/* See fpu.S, this is very similar but to save/restore checkpointed FPRs/VSRs */
+#define __SAVE_32FPRS_VSRS_TRANSACT(n,c,base) \
+BEGIN_FTR_SECTION \
+ b 2f; \
+END_FTR_SECTION_IFSET(CPU_FTR_VSX); \
+ SAVE_32FPRS_TRANSACT(n,base); \
+ b 3f; \
+2: SAVE_32VSRS_TRANSACT(n,c,base); \
+3:
+/* ...and this is just plain borrowed from there. */
+#define __REST_32FPRS_VSRS(n,c,base) \
+BEGIN_FTR_SECTION \
+ b 2f; \
+END_FTR_SECTION_IFSET(CPU_FTR_VSX); \
+ REST_32FPRS(n,base); \
+ b 3f; \
+2: REST_32VSRS(n,c,base); \
+3:
+#else
+#define __SAVE_32FPRS_VSRS_TRANSACT(n,c,base) SAVE_32FPRS_TRANSACT(n, base)
+#define __REST_32FPRS_VSRS(n,c,base) REST_32FPRS(n, base)
+#endif
+#define SAVE_32FPRS_VSRS_TRANSACT(n,c,base) \
+ __SAVE_32FPRS_VSRS_TRANSACT(n,__REG_##c,__REG_##base)
+#define REST_32FPRS_VSRS(n,c,base) \
+ __REST_32FPRS_VSRS(n,__REG_##c,__REG_##base)
+
+/* Stack frame offsets for local variables. */
+#define TM_FRAME_L0 TM_FRAME_SIZE-16
+#define TM_FRAME_L1 TM_FRAME_SIZE-8
+#define STACK_PARAM(x) (48+((x)*8))
+
+
+/* In order to access the TM SPRs, TM must be enabled. So, do so: */
+_GLOBAL(tm_enable)
+ mfmsr r4
+ li r3, MSR_TM >> 32
+ sldi r3, r3, 32
+ and. r0, r4, r3
+ bne 1f
+ or r4, r4, r3
+ mtmsrd r4
+1: blr
+
+_GLOBAL(tm_save_sprs)
+ mfspr r0, SPRN_TFHAR
+ std r0, THREAD_TM_TFHAR(r3)
+ mfspr r0, SPRN_TEXASR
+ std r0, THREAD_TM_TEXASR(r3)
+ mfspr r0, SPRN_TFIAR
+ std r0, THREAD_TM_TFIAR(r3)
+ blr
+
+_GLOBAL(tm_restore_sprs)
+ ld r0, THREAD_TM_TFHAR(r3)
+ mtspr SPRN_TFHAR, r0
+ ld r0, THREAD_TM_TEXASR(r3)
+ mtspr SPRN_TEXASR, r0
+ ld r0, THREAD_TM_TFIAR(r3)
+ mtspr SPRN_TFIAR, r0
+ blr
+
+ /* Passed an 8-bit failure cause as first argument. */
+_GLOBAL(tm_abort)
+ TABORT(R3)
+ blr
+
+
+/* void tm_reclaim(struct thread_struct *thread,
+ * unsigned long orig_msr,
+ * uint8_t cause)
+ *
+ * - Performs a full reclaim. This destroys outstanding
+ * transactions and updates thread->regs.tm_ckpt_* with the
+ * original checkpointed state. Note that thread->regs is
+ * unchanged.
+ * - FP regs are written back to thread->transact_fpr before
+ * reclaiming. These are the transactional (current) versions.
+ *
+ * Purpose is to both abort transactions of, and preserve the state of,
+ * a transactions at a context switch. We preserve/restore both sets of process
+ * state to restore them when the thread's scheduled again. We continue in
+ * userland as though nothing happened, but when the transaction is resumed
+ * they will abort back to the checkpointed state we save out here.
+ */
+_GLOBAL(tm_reclaim)
+ mfcr r6
+ mflr r0
+ std r6, 8(r1)
+ std r0, 16(r1)
+ std r2, 40(r1)
+ stdu r1, -TM_FRAME_SIZE(r1)
+
+ /* We've a struct pt_regs at [r1+STACK_FRAME_OVERHEAD]. */
+
+ std r3, STACK_PARAM(0)(r1)
+ SAVE_NVGPRS(r1)
+
+ /* hard_irq_disable */
+ mfmsr r14
+ li r6, MSR_EE-1
+ addi r6, r6, 1 /* Get around sxt on 0x8000 */
+ andc r15, r14, r6
+ ori r15, r15, MSR_FP
+ oris r15, r15, MSR_VEC@h
+#ifdef CONFIG_VSX
+ BEGIN_FTR_SECTION
+ oris r15,r15, MSR_VSX@h
+ END_FTR_SECTION_IFSET(CPU_FTR_VSX)
+#endif
+ mtmsrd r15
+ std r14, TM_FRAME_L0(r1)
+
+ std r1, PACAR1(r13)
+
+ /* ******************** FPR/VR/VSRs ************
+ * Before reclaiming, capture the current/transactional FPR/VR
+ * versions /if used/.
+ *
+ * (If VSX used, FP and VMX are implied. Or, we don't need to look
+ * at MSR.VSX as copying FP regs if .FP, vector regs if .VMX covers it.)
+ *
+ * We're passed the thread's MSR as parameter 2.
+ *
+ * We enabled VEC/FP/VSX in the msr above, so we can execute these
+ * instructions!
+ */
+ andis. r0, r4, MSR_VEC@h
+ beq dont_backup_vec
+
+ SAVE_32VRS_TRANSACT(0, r6, r3) /* r6 scratch, r3 thread */
+ mfvscr vr0
+ li r6, THREAD_TRANSACT_VSCR
+ stvx vr0, r3, r6
+ mfspr r0, SPRN_VRSAVE
+ std r0, THREAD_TRANSACT_VRSAVE(r3)
+
+dont_backup_vec:
+ andi. r0, r4, MSR_FP
+ beq dont_backup_fp
+
+ SAVE_32FPRS_VSRS_TRANSACT(0, R6, R3) /* r6 scratch, r3 thread */
+
+ mffs fr0
+ stfd fr0,THREAD_TRANSACT_FPSCR(r3)
+
+dont_backup_fp:
+ /* The moment we treclaim, ALL of our GPRs will switch
+ * to user register state. (FPRs, CCR etc. also!)
+ * Use an sprg and a scratch dword in the PACA to shuffle.
+ */
+ TRECLAIM(R5) /* Cause in r5 */
+
+ /* ******************** GPRs ******************** */
+ SET_SCRATCH0(r13)
+ GET_PACA(r13)
+ std r1, PACATMSCRATCH(r13)
+
+ ld r1, PACAR1(r13) /* Get our stack back */
+ /* Temporarily stash checkpointed r0 here to get another
+ * scratch reg for our userland pt_regs pointer:
+ */
+ std r7, GPR7(r1) /* Temporary stash */
+ std r12, GPR12(r1) /* '' '' '' */
+ ld r12, STACK_PARAM(0)(r1) /* Param 0, thread_struct * */
+
+ addi r7, r12, PT_CKPT_REGS /* Thread's ckpt_regs */
+
+ /* Make r7 look like an exception frame so that we
+ * can use the neat GPRx(n) macros. r7 is NOT a pt_regs ptr!
+ */
+ subi r7, r7, STACK_FRAME_OVERHEAD
+
+ /* Sync the userland GPRs 2-12, 14-31 to thread->regs: */
+ SAVE_GPR(0, r7) /* user r0 */
+ SAVE_GPR(2, r7) /* user r2 */
+ SAVE_4GPRS(3, r7) /* user r3-r6 */
+ SAVE_4GPRS(8, r7) /* user r8-r11 */
+ ld r3, PACATMSCRATCH(r13) /* user r1 */
+ ld r4, GPR7(r1) /* user r7 */
+ ld r5, GPR12(r1) /* user r12 */
+ GET_SCRATCH0(6) /* user r13 */
+ std r3, GPR1(r7)
+ std r4, GPR7(r7)
+ std r5, GPR12(r7)
+ std r6, GPR13(r7)
+
+ SAVE_NVGPRS(r7) /* user r14-r31 */
+
+ /* ******************** NIP ******************** */
+ mfspr r3, SPRN_TFHAR
+ std r3, _NIP(r7) /* Returns to failhandler */
+ /* The checkpointed NIP is ignored when rescheduling/rechkpting,
+ * but is used in signal return to 'wind back' to the abort handler.
+ */
+
+ /* ******************** CR,LR,CCR,MSR ********** */
+ mfctr r3
+ mflr r4
+ mfcr r5
+ mfxer r6
+
+ std r3, _CTR(r7)
+ std r4, _LINK(r7)
+ std r5, _CCR(r7)
+ std r6, _XER(r7)
+
+ /* MSR and flags: We don't change CRs, and we don't need to alter
+ * MSR.
+ */
+
+ /* TM regs, incl TEXASR -- these live in thread_struct. Note they've
+ * been updated by the treclaim, to explain to userland the failure
+ * cause (aborted).
+ */
+ mfspr r0, SPRN_TEXASR
+ mfspr r3, SPRN_TFHAR
+ mfspr r4, SPRN_TFIAR
+ std r0, THREAD_TM_TEXASR(r12)
+ std r3, THREAD_TM_TFHAR(r12)
+ std r4, THREAD_TM_TFIAR(r12)
+
+ /* AMR and PPR are checkpointed too, but are unsupported by Linux. */
+
+ /* Restore original MSR/IRQ state & clear TM mode */
+ ld r14, TM_FRAME_L0(r1) /* Orig MSR */
+ li r15, 0
+ rldimi r14, r15, MSR_TS_LG, (63-MSR_TS_LG)-1
+ mtmsrd r14
+
+ REST_NVGPRS(r1)
+
+ addi r1, r1, TM_FRAME_SIZE
+ ld r4, 8(r1)
+ ld r0, 16(r1)
+ mtcr r4
+ mtlr r0
+ ld r2, 40(r1)
+ blr
+
+
+ /* void tm_recheckpoint(struct thread_struct *thread,
+ * unsigned long orig_msr)
+ * - Restore the checkpointed register state saved by tm_reclaim
+ * when we switch_to a process.
+ *
+ * Call with IRQs off, stacks get all out of sync for
+ * some periods in here!
+ */
+_GLOBAL(tm_recheckpoint)
+ mfcr r5
+ mflr r0
+ std r5, 8(r1)
+ std r0, 16(r1)
+ std r2, 40(r1)
+ stdu r1, -TM_FRAME_SIZE(r1)
+
+ /* We've a struct pt_regs at [r1+STACK_FRAME_OVERHEAD].
+ * This is used for backing up the NVGPRs:
+ */
+ SAVE_NVGPRS(r1)
+
+ std r1, PACAR1(r13)
+
+ /* Load complete register state from ts_ckpt* registers */
+
+ addi r7, r3, PT_CKPT_REGS /* Thread's ckpt_regs */
+
+ /* Make r7 look like an exception frame so that we
+ * can use the neat GPRx(n) macros. r7 is now NOT a pt_regs ptr!
+ */
+ subi r7, r7, STACK_FRAME_OVERHEAD
+
+ SET_SCRATCH0(r1)
+
+ mfmsr r6
+ /* R4 = original MSR to indicate whether thread used FP/Vector etc. */
+
+ /* Enable FP/vec in MSR if necessary! */
+ lis r5, MSR_VEC@h
+ ori r5, r5, MSR_FP
+ and. r5, r4, r5
+ beq restore_gprs /* if neither, skip both */
+
+#ifdef CONFIG_VSX
+ BEGIN_FTR_SECTION
+ oris r5, r5, MSR_VSX@h
+ END_FTR_SECTION_IFSET(CPU_FTR_VSX)
+#endif
+ or r5, r6, r5 /* Set MSR.FP+.VSX/.VEC */
+ mtmsr r5
+
+ /* FP and VEC registers: These are recheckpointed from thread.fpr[]
+ * and thread.vr[] respectively. The thread.transact_fpr[] version
+ * is more modern, and will be loaded subsequently by any FPUnavailable
+ * trap.
+ */
+ andis. r0, r4, MSR_VEC@h
+ beq dont_restore_vec
+
+ li r5, THREAD_VSCR
+ lvx vr0, r3, r5
+ mtvscr vr0
+ REST_32VRS(0, r5, r3) /* r5 scratch, r3 THREAD ptr */
+ ld r5, THREAD_VRSAVE(r3)
+ mtspr SPRN_VRSAVE, r5
+
+dont_restore_vec:
+ andi. r0, r4, MSR_FP
+ beq dont_restore_fp
+
+ lfd fr0, THREAD_FPSCR(r3)
+ MTFSF_L(fr0)
+ REST_32FPRS_VSRS(0, R4, R3)
+
+dont_restore_fp:
+ mtmsr r6 /* FP/Vec off again! */
+
+restore_gprs:
+ /* ******************** CR,LR,CCR,MSR ********** */
+ ld r3, _CTR(r7)
+ ld r4, _LINK(r7)
+ ld r5, _CCR(r7)
+ ld r6, _XER(r7)
+
+ mtctr r3
+ mtlr r4
+ mtcr r5
+ mtxer r6
+
+ /* MSR and flags: We don't change CRs, and we don't need to alter
+ * MSR.
+ */
+
+ REST_4GPRS(0, r7) /* GPR0-3 */
+ REST_GPR(4, r7) /* GPR4-6 */
+ REST_GPR(5, r7)
+ REST_GPR(6, r7)
+ REST_4GPRS(8, r7) /* GPR8-11 */
+ REST_2GPRS(12, r7) /* GPR12-13 */
+
+ REST_NVGPRS(r7) /* GPR14-31 */
+
+ ld r7, GPR7(r7) /* GPR7 */
+
+ /* Commit register state as checkpointed state: */
+ TRECHKPT
+
+ /* Our transactional state has now changed.
+ *
+ * Now just get out of here. Transactional (current) state will be
+ * updated once restore is called on the return path in the _switch-ed
+ * -to process.
+ */
+
+ GET_PACA(r13)
+ GET_SCRATCH0(r1)
+
+ REST_NVGPRS(r1)
+
+ addi r1, r1, TM_FRAME_SIZE
+ ld r4, 8(r1)
+ ld r0, 16(r1)
+ mtcr r4
+ mtlr r0
+ ld r2, 40(r1)
+ blr
+
+ /* ****************************************************************** */
--
1.7.10.4
^ permalink raw reply related
* [PATCH 08/17] powerpc: Add FP/VSX and VMX register load functions for transactional memory
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
This adds functions to restore the state of the FP/VSX registers from
what's stored in the thread_struct. Two version for FP/VSX are required
since one restores them from transactional/checkpoint side of the
thread_struct and the other from the speculated side.
Similar functions are added for VMX registers.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/kernel/fpu.S | 54 ++++++++++++++++++++++++++++++++++++++++++
arch/powerpc/kernel/vector.S | 51 +++++++++++++++++++++++++++++++++++++++
2 files changed, 105 insertions(+)
diff --git a/arch/powerpc/kernel/fpu.S b/arch/powerpc/kernel/fpu.S
index adb1551..6ab0e87 100644
--- a/arch/powerpc/kernel/fpu.S
+++ b/arch/powerpc/kernel/fpu.S
@@ -62,6 +62,60 @@ END_FTR_SECTION_IFSET(CPU_FTR_VSX); \
__REST_32FPVSRS_TRANSACT(n,__REG_##c,__REG_##base)
#define SAVE_32FPVSRS(n,c,base) __SAVE_32FPVSRS(n,__REG_##c,__REG_##base)
+#ifdef CONFIG_TRANSACTIONAL_MEM
+/*
+ * Wrapper to call load_up_fpu from C.
+ * void do_load_up_fpu(struct pt_regs *regs);
+ */
+_GLOBAL(do_load_up_fpu)
+ mflr r0
+ std r0, 16(r1)
+ stdu r1, -112(r1)
+
+ subi r6, r3, STACK_FRAME_OVERHEAD
+ /* load_up_fpu expects r12=MSR, r13=PACA, and returns
+ * with r12 = new MSR.
+ */
+ ld r12,_MSR(r6)
+ GET_PACA(r13)
+
+ bl load_up_fpu
+ std r12,_MSR(r6)
+
+ ld r0, 112+16(r1)
+ addi r1, r1, 112
+ mtlr r0
+ blr
+
+
+/* void do_load_up_fpu(struct thread_struct *thread)
+ *
+ * This is similar to load_up_fpu but for the transactional version of the FP
+ * register set. It doesn't mess with the task MSR or valid flags.
+ * Furthermore, we don't do lazy FP with TM currently.
+ */
+_GLOBAL(do_load_up_transact_fpu)
+ mfmsr r6
+ ori r5,r6,MSR_FP
+#ifdef CONFIG_VSX
+BEGIN_FTR_SECTION
+ oris r5,r5,MSR_VSX@h
+END_FTR_SECTION_IFSET(CPU_FTR_VSX)
+#endif
+ SYNC
+ MTMSRD(r5)
+
+ lfd fr0,THREAD_TRANSACT_FPSCR(r3)
+ MTFSF_L(fr0)
+ REST_32FPVSRS_TRANSACT(0, R4, R3)
+
+ /* FP/VSX off again */
+ MTMSRD(r6)
+ SYNC
+
+ blr
+#endif /* CONFIG_TRANSACTIONAL_MEM */
+
/*
* This task wants to use the FPU now.
* On UP, disable FP for the task which had the FPU previously,
diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S
index e830289..330fc8c 100644
--- a/arch/powerpc/kernel/vector.S
+++ b/arch/powerpc/kernel/vector.S
@@ -7,6 +7,57 @@
#include <asm/page.h>
#include <asm/ptrace.h>
+#ifdef CONFIG_TRANSACTIONAL_MEM
+/*
+ * Wrapper to call load_up_altivec from C.
+ * void do_load_up_altivec(struct pt_regs *regs);
+ */
+_GLOBAL(do_load_up_altivec)
+ mflr r0
+ std r0, 16(r1)
+ stdu r1, -112(r1)
+
+ subi r6, r3, STACK_FRAME_OVERHEAD
+ /* load_up_altivec expects r12=MSR, r13=PACA, and returns
+ * with r12 = new MSR.
+ */
+ ld r12,_MSR(r6)
+ GET_PACA(r13)
+ bl load_up_altivec
+ std r12,_MSR(r6)
+
+ ld r0, 112+16(r1)
+ addi r1, r1, 112
+ mtlr r0
+ blr
+
+/* void do_load_up_altivec(struct thread_struct *thread)
+ *
+ * This is similar to load_up_altivec but for the transactional version of the
+ * vector regs. It doesn't mess with the task MSR or valid flags.
+ * Furthermore, VEC laziness is not supported with TM currently.
+ */
+_GLOBAL(do_load_up_transact_altivec)
+ mfmsr r6
+ oris r5,r6,MSR_VEC@h
+ MTMSRD(r5)
+ isync
+
+ li r4,1
+ stw r4,THREAD_USED_VR(r3)
+
+ li r10,THREAD_TRANSACT_VSCR
+ lvx vr0,r10,r3
+ mtvscr vr0
+ REST_32VRS_TRANSACT(0,r4,r3)
+
+ /* Disable VEC again. */
+ MTMSRD(r6)
+ isync
+
+ blr
+#endif
+
/*
* load_up_altivec(unused, unused, tsk)
* Disable VMX for the task which had it previously,
--
1.7.10.4
^ permalink raw reply related
* [PATCH 09/17] powerpc: Add reclaim and recheckpoint functions for context switching transactional memory processes
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
When we switch out a task, we need to save both the checkpointed and the
speculated state into the thread struct.
Similarly when we are switching in a task we need to load both the checkpointed
and speculated state. If the task was using FP, we non-lazily reload both the
original and the speculative FP register states. This is because the kernel
doesn't see if/when a TM rollback occurs, so if we take an FP unavoidable
later, we are unable to determine which set of FP regs need to be restored.
This simply adds these functions. It doesn't hook them into the existing code
yet.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/kernel/process.c | 112 +++++++++++++++++++++++++++++++++++++++++
1 file changed, 112 insertions(+)
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index cb11fed..14fabd1 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -50,6 +50,7 @@
#include <asm/runlatch.h>
#include <asm/syscalls.h>
#include <asm/switch_to.h>
+#include <asm/tm.h>
#include <asm/debug.h>
#ifdef CONFIG_PPC64
#include <asm/firmware.h>
@@ -462,6 +463,117 @@ static inline bool hw_brk_match(struct arch_hw_breakpoint *a,
return false;
return true;
}
+#ifdef CONFIG_TRANSACTIONAL_MEM
+static inline void tm_reclaim_task(struct task_struct *tsk)
+{
+ /* We have to work out if we're switching from/to a task that's in the
+ * middle of a transaction.
+ *
+ * In switching we need to maintain a 2nd register state as
+ * oldtask->thread.ckpt_regs. We tm_reclaim(oldproc); this saves the
+ * checkpointed (tbegin) state in ckpt_regs and saves the transactional
+ * (current) FPRs into oldtask->thread.transact_fpr[].
+ *
+ * We also context switch (save) TFHAR/TEXASR/TFIAR in here.
+ */
+ struct thread_struct *thr = &tsk->thread;
+
+ if (!thr->regs)
+ return;
+
+ if (!MSR_TM_ACTIVE(thr->regs->msr))
+ goto out_and_saveregs;
+
+ /* Stash the original thread MSR, as giveup_fpu et al will
+ * modify it. We hold onto it to see whether the task used
+ * FP & vector regs.
+ */
+ thr->tm_orig_msr = thr->regs->msr;
+
+ TM_DEBUG("--- tm_reclaim on pid %d (NIP=%lx, "
+ "ccr=%lx, msr=%lx, trap=%lx)\n",
+ tsk->pid, thr->regs->nip,
+ thr->regs->ccr, thr->regs->msr,
+ thr->regs->trap);
+
+ tm_reclaim(thr, thr->regs->msr, TM_CAUSE_RESCHED);
+
+ TM_DEBUG("--- tm_reclaim on pid %d complete\n",
+ tsk->pid);
+
+out_and_saveregs:
+ /* Always save the regs here, even if a transaction's not active.
+ * This context-switches a thread's TM info SPRs. We do it here to
+ * be consistent with the restore path (in recheckpoint) which
+ * cannot happen later in _switch().
+ */
+ tm_save_sprs(thr);
+}
+
+static inline void __maybe_unused tm_recheckpoint_new_task(struct task_struct *new)
+{
+ unsigned long msr;
+
+ if (!cpu_has_feature(CPU_FTR_TM))
+ return;
+
+ /* Recheckpoint the registers of the thread we're about to switch to.
+ *
+ * If the task was using FP, we non-lazily reload both the original and
+ * the speculative FP register states. This is because the kernel
+ * doesn't see if/when a TM rollback occurs, so if we take an FP
+ * unavoidable later, we are unable to determine which set of FP regs
+ * need to be restored.
+ */
+ if (!new->thread.regs)
+ return;
+
+ /* The TM SPRs are restored here, so that TEXASR.FS can be set
+ * before the trecheckpoint and no explosion occurs.
+ */
+ tm_restore_sprs(&new->thread);
+
+ if (!MSR_TM_ACTIVE(new->thread.regs->msr))
+ return;
+ msr = new->thread.tm_orig_msr;
+ /* Recheckpoint to restore original checkpointed register state. */
+ TM_DEBUG("*** tm_recheckpoint of pid %d "
+ "(new->msr 0x%lx, new->origmsr 0x%lx)\n",
+ new->pid, new->thread.regs->msr, msr);
+
+ /* This loads the checkpointed FP/VEC state, if used */
+ tm_recheckpoint(&new->thread, msr);
+
+ /* This loads the speculative FP/VEC state, if used */
+ if (msr & MSR_FP) {
+ do_load_up_transact_fpu(&new->thread);
+ new->thread.regs->msr |=
+ (MSR_FP | new->thread.fpexc_mode);
+ }
+ if (msr & MSR_VEC) {
+ do_load_up_transact_altivec(&new->thread);
+ new->thread.regs->msr |= MSR_VEC;
+ }
+ /* We may as well turn on VSX too since all the state is restored now */
+ if (msr & MSR_VSX)
+ new->thread.regs->msr |= MSR_VSX;
+
+ TM_DEBUG("*** tm_recheckpoint of pid %d complete "
+ "(kernel msr 0x%lx)\n",
+ new->pid, mfmsr());
+}
+
+static inline void __switch_to_tm(struct task_struct *prev)
+{
+ if (cpu_has_feature(CPU_FTR_TM)) {
+ tm_enable();
+ tm_reclaim_task(prev);
+ }
+}
+#else
+#define tm_recheckpoint_new_task(new)
+#define __switch_to_tm(prev)
+#endif /* CONFIG_TRANSACTIONAL_MEM */
struct task_struct *__switch_to(struct task_struct *prev,
struct task_struct *new)
--
1.7.10.4
^ permalink raw reply related
* [PATCH 10/17] powerpc: Add transactional memory unavaliable execption handler
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
These should never happen since we always turn on MSR TM when in userspace. We
don't do lazy TM.
Hence if we hit this, we barf and kill the task as something's gone horribly
wrong.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/kernel/exceptions-64s.S | 19 +++++++++++++++++++
arch/powerpc/kernel/traps.c | 21 +++++++++++++++++++++
2 files changed, 40 insertions(+)
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index 7a1c87c..d0cc657 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -313,6 +313,9 @@ vsx_unavailable_pSeries_1:
. = 0xf40
b vsx_unavailable_pSeries
+ . = 0xf60
+ b tm_unavailable_pSeries
+
#ifdef CONFIG_CBE_RAS
STD_EXCEPTION_HV(0x1200, 0x1202, cbe_system_error)
KVM_HANDLER_SKIP(PACA_EXGEN, EXC_HV, 0x1202)
@@ -528,6 +531,8 @@ ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
KVM_HANDLER_PR(PACA_EXGEN, EXC_STD, 0xf20)
STD_EXCEPTION_PSERIES(., 0xf40, vsx_unavailable)
KVM_HANDLER_PR(PACA_EXGEN, EXC_STD, 0xf40)
+ STD_EXCEPTION_PSERIES(., 0xf60, tm_unavailable)
+ KVM_HANDLER_PR(PACA_EXGEN, EXC_STD, 0xf60)
/*
* An interrupt came in while soft-disabled. We set paca->irq_happened, then:
@@ -817,6 +822,10 @@ vsx_unavailable_relon_pSeries_1:
. = 0x4f40
b vsx_unavailable_relon_pSeries
+tm_unavailable_relon_pSeries_1:
+ . = 0x4f60
+ b tm_unavailable_relon_pSeries
+
#ifdef CONFIG_CBE_RAS
STD_RELON_EXCEPTION_HV(0x5200, 0x1202, cbe_system_error)
#endif /* CONFIG_CBE_RAS */
@@ -1177,6 +1186,15 @@ END_FTR_SECTION_IFSET(CPU_FTR_VSX)
b .ret_from_except
.align 7
+ .globl tm_unavailable_common
+tm_unavailable_common:
+ EXCEPTION_PROLOG_COMMON(0xf60, PACA_EXGEN)
+ bl .save_nvgprs
+ addi r3,r1,STACK_FRAME_OVERHEAD
+ bl .tm_unavailable_exception
+ b .ret_from_except
+
+ .align 7
.globl __end_handlers
__end_handlers:
@@ -1195,6 +1213,7 @@ __end_handlers:
STD_RELON_EXCEPTION_PSERIES(., 0xf00, performance_monitor)
STD_RELON_EXCEPTION_PSERIES(., 0xf20, altivec_unavailable)
STD_RELON_EXCEPTION_PSERIES(., 0xf40, vsx_unavailable)
+ STD_RELON_EXCEPTION_PSERIES(., 0xf60, tm_unavailable)
#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
/*
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index 2a62f54..dda0517 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -1171,6 +1171,27 @@ void vsx_unavailable_exception(struct pt_regs *regs)
die("Unrecoverable VSX Unavailable Exception", regs, SIGABRT);
}
+void tm_unavailable_exception(struct pt_regs *regs)
+{
+ /* We restore the interrupt state now */
+ if (!arch_irq_disabled_regs(regs))
+ local_irq_enable();
+
+ /* Currently we never expect a TMU exception. Catch
+ * this and kill the process!
+ */
+ printk(KERN_EMERG "Unexpected TM unavailable exception at %lx "
+ "(msr %lx)\n",
+ regs->nip, regs->msr);
+
+ if (user_mode(regs)) {
+ _exception(SIGILL, regs, ILL_ILLOPC, regs->nip);
+ return;
+ }
+
+ die("Unexpected TM unavailable exception", regs, SIGABRT);
+}
+
void performance_monitor_exception(struct pt_regs *regs)
{
__get_cpu_var(irq_stat).pmu_irqs++;
--
1.7.10.4
^ permalink raw reply related
* [PATCH 11/17] powerpc: Assembler routines for FP/VSX/VMX unavailable during a transaction
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
We do lazy FP but not lazy TM (ie. userspace starts with MSR TM=1 FP=0). Hence
if userspace does an FP instruction during a transaction, we'll take an
fp unavailable exception.
This adds functions needed to handle this case. We have to inject the current
FP state into the checkpoint so that the hardware can decide what to do with
the transaction. We can't inject only the FP so we have to do a full treclaim
and recheckpoint to inject just the FP state. This will cause the transaction
to be marked as aborted by the hardware.
This just add the routines needed to do this for FP, VMX and VSX. It doesn't
hook them into the rest of the code yet.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/kernel/traps.c | 95 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 95 insertions(+)
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index dda0517..311770d 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -58,6 +58,7 @@
#include <asm/rio.h>
#include <asm/fadump.h>
#include <asm/switch_to.h>
+#include <asm/tm.h>
#include <asm/debug.h>
#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC)
@@ -1192,6 +1193,100 @@ void tm_unavailable_exception(struct pt_regs *regs)
die("Unexpected TM unavailable exception", regs, SIGABRT);
}
+#ifdef CONFIG_TRANSACTIONAL_MEM
+
+extern void do_load_up_fpu(struct pt_regs *regs);
+
+void fp_unavailable_tm(struct pt_regs *regs)
+{
+ /* Note: This does not handle any kind of FP laziness. */
+
+ /* We restore the interrupt state now */
+ if (!arch_irq_disabled_regs(regs))
+ local_irq_enable();
+
+ TM_DEBUG("FP Unavailable trap whilst transactional at 0x%lx, MSR=%lx\n",
+ regs->nip, regs->msr);
+ tm_enable();
+
+ /* We can only have got here if the task started using FP after
+ * beginning the transaction. So, the transactional regs are just a
+ * copy of the checkpointed ones. But, we still need to recheckpoint
+ * as we're enabling FP for the process; it will return, abort the
+ * transaction, and probably retry but now with FP enabled. So the
+ * checkpointed FP registers need to be loaded.
+ */
+ tm_reclaim(¤t->thread, current->thread.regs->msr,
+ TM_CAUSE_FAC_UNAV);
+ /* Reclaim didn't save out any FPRs to transact_fprs. */
+
+ /* Enable FP for the task: */
+ regs->msr |= (MSR_FP | current->thread.fpexc_mode);
+
+ /* This loads and recheckpoints the FP registers from
+ * thread.fpr[]. They will remain in registers after the
+ * checkpoint so we don't need to reload them after.
+ */
+ tm_recheckpoint(¤t->thread, regs->msr);
+}
+
+#ifdef CONFIG_ALTIVEC
+extern void do_load_up_altivec(struct pt_regs *regs);
+
+void altivec_unavailable_tm(struct pt_regs *regs)
+{
+ /* See the comments in fp_unavailable_tm(). This function operates
+ * the same way.
+ */
+
+ /* We restore the interrupt state now */
+ if (!arch_irq_disabled_regs(regs))
+ local_irq_enable();
+
+ TM_DEBUG("Vector Unavailable trap whilst transactional at 0x%lx,"
+ "MSR=%lx\n",
+ regs->nip, regs->msr);
+ tm_enable();
+ tm_reclaim(¤t->thread, current->thread.regs->msr,
+ TM_CAUSE_FAC_UNAV);
+ regs->msr |= MSR_VEC;
+ tm_recheckpoint(¤t->thread, regs->msr);
+ current->thread.used_vr = 1;
+}
+#endif
+
+#ifdef CONFIG_VSX
+void vsx_unavailable_tm(struct pt_regs *regs)
+{
+ /* See the comments in fp_unavailable_tm(). This works similarly,
+ * though we're loading both FP and VEC registers in here.
+ *
+ * If FP isn't in use, load FP regs. If VEC isn't in use, load VEC
+ * regs. Either way, set MSR_VSX.
+ */
+
+ /* We restore the interrupt state now */
+ if (!arch_irq_disabled_regs(regs))
+ local_irq_enable();
+
+ TM_DEBUG("VSX Unavailable trap whilst transactional at 0x%lx,"
+ "MSR=%lx\n",
+ regs->nip, regs->msr);
+
+ tm_enable();
+ /* This reclaims FP and/or VR regs if they're already enabled */
+ tm_reclaim(¤t->thread, current->thread.regs->msr,
+ TM_CAUSE_FAC_UNAV);
+
+ regs->msr |= MSR_VEC | MSR_FP | current->thread.fpexc_mode |
+ MSR_VSX;
+ /* This loads & recheckpoints FP and VRs. */
+ tm_recheckpoint(¤t->thread, regs->msr);
+ current->thread.used_vsr = 1;
+}
+#endif
+#endif /* CONFIG_TRANSACTIONAL_MEM */
+
void performance_monitor_exception(struct pt_regs *regs)
{
__get_cpu_var(irq_stat).pmu_irqs++;
--
1.7.10.4
^ permalink raw reply related
* [PATCH 12/17] powerpc: Hook in new transactional memory code
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
This hooks the new transactional memory code into context switching, FP/VMX/VMX
unavailable and exception return.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/kernel/exceptions-64s.S | 48 ++++++++++++++++++++++++++++++++--
arch/powerpc/kernel/fpu.S | 1 -
arch/powerpc/kernel/process.c | 15 +++++++++--
arch/powerpc/kernel/traps.c | 32 +++++++++++++++++++++++
arch/powerpc/kernel/vector.S | 1 -
arch/powerpc/mm/hash_utils_64.c | 13 +++++++++
6 files changed, 104 insertions(+), 6 deletions(-)
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index d0cc657..c4a20d1 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -1147,9 +1147,24 @@ fp_unavailable_common:
addi r3,r1,STACK_FRAME_OVERHEAD
bl .kernel_fp_unavailable_exception
BUG_OPCODE
-1: bl .load_up_fpu
+1:
+#ifdef CONFIG_TRANSACTIONAL_MEM
+BEGIN_FTR_SECTION
+ srdi r0, r12, MSR_TS_LG
+ andi. r0, r0, 3
+ bne- 2f
+END_FTR_SECTION_IFSET(CPU_FTR_TM)
+#endif
+ bl .load_up_fpu
+ std r12,_MSR(r1)
b fast_exception_return
-
+#ifdef CONFIG_TRANSACTIONAL_MEM
+2: /* User process was in a transaction */
+ bl .save_nvgprs
+ addi r3,r1,STACK_FRAME_OVERHEAD
+ bl .fp_unavailable_tm
+ b .ret_from_except
+#endif
.align 7
.globl altivec_unavailable_common
altivec_unavailable_common:
@@ -1157,8 +1172,23 @@ altivec_unavailable_common:
#ifdef CONFIG_ALTIVEC
BEGIN_FTR_SECTION
beq 1f
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ BEGIN_FTR_SECTION_NESTED(69)
+ srdi r0, r12, MSR_TS_LG
+ andi. r0, r0, 3
+ bne- 2f
+ END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
+#endif
bl .load_up_altivec
+ std r12,_MSR(r1)
b fast_exception_return
+#ifdef CONFIG_TRANSACTIONAL_MEM
+2: /* User process was in a transaction */
+ bl .save_nvgprs
+ addi r3,r1,STACK_FRAME_OVERHEAD
+ bl .altivec_unavailable_tm
+ b .ret_from_except
+#endif
1:
END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
#endif
@@ -1175,7 +1205,21 @@ vsx_unavailable_common:
#ifdef CONFIG_VSX
BEGIN_FTR_SECTION
beq 1f
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ BEGIN_FTR_SECTION_NESTED(69)
+ srdi r0, r12, MSR_TS_LG
+ andi. r0, r0, 3
+ bne- 2f
+ END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
+#endif
b .load_up_vsx
+#ifdef CONFIG_TRANSACTIONAL_MEM
+2: /* User process was in a transaction */
+ bl .save_nvgprs
+ addi r3,r1,STACK_FRAME_OVERHEAD
+ bl .vsx_unavailable_tm
+ b .ret_from_except
+#endif
1:
END_FTR_SECTION_IFSET(CPU_FTR_VSX)
#endif
diff --git a/arch/powerpc/kernel/fpu.S b/arch/powerpc/kernel/fpu.S
index 6ab0e87..08b6a12f 100644
--- a/arch/powerpc/kernel/fpu.S
+++ b/arch/powerpc/kernel/fpu.S
@@ -170,7 +170,6 @@ END_FTR_SECTION_IFSET(CPU_FTR_VSX)
lwz r4,THREAD_FPEXC_MODE(r5)
ori r12,r12,MSR_FP
or r12,r12,r4
- std r12,_MSR(r1)
#endif
lfd fr0,THREAD_FPSCR(r5)
MTFSF_L(fr0)
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index 14fabd1..bccc81b 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -510,7 +510,7 @@ out_and_saveregs:
tm_save_sprs(thr);
}
-static inline void __maybe_unused tm_recheckpoint_new_task(struct task_struct *new)
+static inline void tm_recheckpoint_new_task(struct task_struct *new)
{
unsigned long msr;
@@ -585,6 +585,8 @@ struct task_struct *__switch_to(struct task_struct *prev,
struct ppc64_tlb_batch *batch;
#endif
+ __switch_to_tm(prev);
+
#ifdef CONFIG_SMP
/* avoid complexity of lazy save/restore of fpu
* by just saving it every time we switch out if
@@ -700,6 +702,9 @@ struct task_struct *__switch_to(struct task_struct *prev,
* of sync. Hard disable here.
*/
hard_irq_disable();
+
+ tm_recheckpoint_new_task(new);
+
last = _switch(old_thread, new_thread);
#ifdef CONFIG_PPC_BOOK3S_64
@@ -1075,7 +1080,6 @@ void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
regs->msr = MSR_USER32;
}
#endif
-
discard_lazy_cpu_state();
#ifdef CONFIG_VSX
current->thread.used_vsr = 0;
@@ -1095,6 +1099,13 @@ void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
current->thread.spefscr = 0;
current->thread.used_spe = 0;
#endif /* CONFIG_SPE */
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ if (cpu_has_feature(CPU_FTR_TM))
+ regs->msr |= MSR_TM;
+ current->thread.tm_tfhar = 0;
+ current->thread.tm_texasr = 0;
+ current->thread.tm_tfiar = 0;
+#endif /* CONFIG_TRANSACTIONAL_MEM */
}
#define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index 311770d..486712b 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -1032,6 +1032,38 @@ void __kprobes program_check_exception(struct pt_regs *regs)
_exception(SIGTRAP, regs, TRAP_BRKPT, regs->nip);
return;
}
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ if (reason & REASON_TM) {
+ /* This is a TM "Bad Thing Exception" program check.
+ * This occurs when:
+ * - An rfid/hrfid/mtmsrd attempts to cause an illegal
+ * transition in TM states.
+ * - A trechkpt is attempted when transactional.
+ * - A treclaim is attempted when non transactional.
+ * - A tend is illegally attempted.
+ * - writing a TM SPR when transactional.
+ */
+ if (!user_mode(regs) &&
+ report_bug(regs->nip, regs) == BUG_TRAP_TYPE_WARN) {
+ regs->nip += 4;
+ return;
+ }
+ /* If usermode caused this, it's done something illegal and
+ * gets a SIGILL slap on the wrist. We call it an illegal
+ * operand to distinguish from the instruction just being bad
+ * (e.g. executing a 'tend' on a CPU without TM!); it's an
+ * illegal /placement/ of a valid instruction.
+ */
+ if (user_mode(regs)) {
+ _exception(SIGILL, regs, ILL_ILLOPN, regs->nip);
+ return;
+ } else {
+ printk(KERN_EMERG "Unexpected TM Bad Thing exception "
+ "at %lx (msr 0x%x)\n", regs->nip, reason);
+ die("Unrecoverable exception", regs, SIGABRT);
+ }
+ }
+#endif
/* We restore the interrupt state now */
if (!arch_irq_disabled_regs(regs))
diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S
index 330fc8c..9b47306 100644
--- a/arch/powerpc/kernel/vector.S
+++ b/arch/powerpc/kernel/vector.S
@@ -123,7 +123,6 @@ _GLOBAL(load_up_altivec)
ld r4,PACACURRENT(r13)
addi r5,r4,THREAD /* Get THREAD */
oris r12,r12,MSR_VEC@h
- std r12,_MSR(r1)
#endif
li r4,1
li r10,THREAD_VSCR
diff --git a/arch/powerpc/mm/hash_utils_64.c b/arch/powerpc/mm/hash_utils_64.c
index 3a292be..03a4c4a 100644
--- a/arch/powerpc/mm/hash_utils_64.c
+++ b/arch/powerpc/mm/hash_utils_64.c
@@ -55,6 +55,7 @@
#include <asm/code-patching.h>
#include <asm/fadump.h>
#include <asm/firmware.h>
+#include <asm/tm.h>
#ifdef DEBUG
#define DBG(fmt...) udbg_printf(fmt)
@@ -1171,6 +1172,18 @@ void flush_hash_page(unsigned long vpn, real_pte_t pte, int psize, int ssize,
DBG_LOW(" sub %ld: hash=%lx, hidx=%lx\n", index, slot, hidx);
ppc_md.hpte_invalidate(slot, vpn, psize, ssize, local);
} pte_iterate_hashed_end();
+
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ /* Transactions are not aborted by tlbiel, only tlbie.
+ * Without, syncing a page back to a block device w/ PIO could pick up
+ * transactional data (bad!) so we force an abort here. Before the
+ * sync the page will be made read-only, which will flush_hash_page.
+ */
+ if (local && cpu_has_feature(CPU_FTR_TM)) {
+ tm_enable();
+ tm_abort(TM_CAUSE_TLBI);
+ }
+#endif
}
void flush_hash_range(unsigned long number, int local)
--
1.7.10.4
^ permalink raw reply related
* [PATCH 13/17] powerpc: Add new transactional memory state to the signal context
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
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>
---
arch/powerpc/include/asm/reg.h | 1 +
arch/powerpc/kernel/signal.h | 8 +
arch/powerpc/kernel/signal_32.c | 500 ++++++++++++++++++++++++++++++++++++++-
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 d19581ae4..41894dc 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 e00acb4..ec84c90 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 804e323..9dc2472 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_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_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_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_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_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_TRANSACTIONAL_MEM */
#endif
/*
@@ -483,6 +562,156 @@ static int save_user_regs(struct pt_regs *regs, struct mcontext __user *frame,
return 0;
}
+#ifdef CONFIG_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_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_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_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_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_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_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_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_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_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 1ca045d..5d349bd 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_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_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_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_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_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_mcontext,
+ &uc_transact->uc_mcontext))
+ goto badframe;
+ }
+ else
+ /* Fall through, for non-TM restore */
+#endif
if (restore_sigcontext(regs, NULL, 1, &uc->uc_mcontext))
goto badframe;
@@ -415,19 +723,42 @@ int handle_rt_signal64(int signr, struct k_sigaction *ka, siginfo_t *info,
/* Create the ucontext. */
err |= __put_user(0, &frame->uc.uc_flags);
- err |= __put_user(0, &frame->uc.uc_link);
err |= __put_user(current->sas_ss_sp, &frame->uc.uc_stack.ss_sp);
err |= __put_user(sas_ss_flags(regs->gpr[1]),
&frame->uc.uc_stack.ss_flags);
err |= __put_user(current->sas_ss_size, &frame->uc.uc_stack.ss_size);
- err |= setup_sigcontext(&frame->uc.uc_mcontext, regs, signr, NULL,
- (unsigned long)ka->sa.sa_handler, 1);
+#ifdef CONFIG_TRANSACTIONAL_MEM
+ if (MSR_TM_ACTIVE(regs->msr)) {
+ /* The ucontext_t passed to userland points to the second
+ * ucontext_t (for transactional state) with its uc_link ptr.
+ */
+ err |= __put_user(&frame->uc_transact, &frame->uc.uc_link);
+ err |= setup_tm_sigcontexts(&frame->uc.uc_mcontext,
+ &frame->uc_transact.uc_mcontext,
+ regs, signr,
+ NULL,
+ (unsigned long)ka->sa.sa_handler);
+ } else
+#endif
+ {
+ err |= __put_user(0, &frame->uc.uc_link);
+ err |= setup_sigcontext(&frame->uc.uc_mcontext, regs, signr,
+ NULL, (unsigned long)ka->sa.sa_handler,
+ 1);
+ }
err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set));
if (err)
goto badframe;
/* Make sure signal handler doesn't get spurious FP exceptions */
current->thread.fpscr.val = 0;
+#ifdef CONFIG_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
/* Set up to return from userspace. */
if (vdso64_rt_sigtramp && current->mm->context.vdso_base) {
--
1.7.10.4
^ permalink raw reply related
* [PATCH 14/17] powerpc: Add transactional memory to POWER8 cpu features
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/include/asm/cputable.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/powerpc/include/asm/cputable.h b/arch/powerpc/include/asm/cputable.h
index 71a498b..5ab77f4 100644
--- a/arch/powerpc/include/asm/cputable.h
+++ b/arch/powerpc/include/asm/cputable.h
@@ -421,7 +421,7 @@ extern const char *powerpc_base_platform;
CPU_FTR_DSCR | CPU_FTR_SAO | \
CPU_FTR_STCX_CHECKS_ADDRESS | CPU_FTR_POPCNTB | CPU_FTR_POPCNTD | \
CPU_FTR_ICSWX | CPU_FTR_CFAR | CPU_FTR_HVMODE | CPU_FTR_VMX_COPY | \
- CPU_FTR_DBELL | CPU_FTR_HAS_PPR | CPU_FTR_DAWR)
+ CPU_FTR_DBELL | CPU_FTR_HAS_PPR | CPU_FTR_DAWR | CPU_FTR_TM_COMP)
#define CPU_FTRS_CELL (CPU_FTR_USE_TB | CPU_FTR_LWSYNC | \
CPU_FTR_PPCAS_ARCH_V2 | CPU_FTR_CTRL | \
CPU_FTR_ALTIVEC_COMP | CPU_FTR_MMCRA | CPU_FTR_SMT | \
--
1.7.10.4
^ permalink raw reply related
* [PATCH 15/17] powerpc: Add config option for transactional memory
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Kconfig option for transactional memory on powerpc.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/Kconfig | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 684fa64..349ed12 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -309,6 +309,14 @@ config MATH_EMULATION
unit, which will allow programs that use floating-point
instructions to run.
+config TRANSACTIONAL_MEM
+ bool "Transactional Memory support"
+ depends on PPC64
+ depends on SMP
+ default n
+ ---help---
+ Support user-mode Transactional Memory.
+
config 8XX_MINIMAL_FPEMU
bool "Minimal math emulation for 8xx"
depends on 8xx && !MATH_EMULATION
--
1.7.10.4
^ permalink raw reply related
* [PATCH 16/17] powerpc: Add transactional memory to pseries and ppc64 defconfigs
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
arch/powerpc/configs/ppc64_defconfig | 2 ++
arch/powerpc/configs/pseries_defconfig | 2 ++
2 files changed, 4 insertions(+)
diff --git a/arch/powerpc/configs/ppc64_defconfig b/arch/powerpc/configs/ppc64_defconfig
index 6cbdeb4..01e2e86 100644
--- a/arch/powerpc/configs/ppc64_defconfig
+++ b/arch/powerpc/configs/ppc64_defconfig
@@ -49,6 +49,8 @@ CONFIG_CPU_FREQ_GOV_USERSPACE=y
CONFIG_CPU_FREQ_PMAC64=y
CONFIG_HZ_100=y
CONFIG_BINFMT_MISC=m
+CONFIG_TRANSACTIONAL_MEM=y
+CONFIG_HOTPLUG_CPU=y
CONFIG_KEXEC=y
CONFIG_IRQ_ALL_CPUS=y
CONFIG_MEMORY_HOTREMOVE=y
diff --git a/arch/powerpc/configs/pseries_defconfig b/arch/powerpc/configs/pseries_defconfig
index 04fa7f2..a50484f 100644
--- a/arch/powerpc/configs/pseries_defconfig
+++ b/arch/powerpc/configs/pseries_defconfig
@@ -43,6 +43,8 @@ CONFIG_RTAS_FLASH=m
CONFIG_IBMEBUS=y
CONFIG_HZ_100=y
CONFIG_BINFMT_MISC=m
+CONFIG_TRANSACTIONAL_MEM=y
+CONFIG_HOTPLUG_CPU=y
CONFIG_KEXEC=y
CONFIG_IRQ_ALL_CPUS=y
CONFIG_MEMORY_HOTPLUG=y
--
1.7.10.4
^ permalink raw reply related
* [PATCH 17/17] powerpc: Documentation for transactional memory on powerpc
From: Michael Neuling @ 2013-01-18 5:48 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: Michael Neuling, linuxppc-dev, Matt Evans
In-Reply-To: <1358488117-17363-1-git-send-email-mikey@neuling.org>
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Michael Neuling <mikey@neuling.org>
---
Documentation/powerpc/transactional_memory.txt | 175 ++++++++++++++++++++++++
1 file changed, 175 insertions(+)
create mode 100644 Documentation/powerpc/transactional_memory.txt
diff --git a/Documentation/powerpc/transactional_memory.txt b/Documentation/powerpc/transactional_memory.txt
new file mode 100644
index 0000000..c907be4
--- /dev/null
+++ b/Documentation/powerpc/transactional_memory.txt
@@ -0,0 +1,175 @@
+Transactional Memory support
+============================
+
+POWER kernel support for this feature is currently limited to supporting
+its use by user programs. It is not currently used by the kernel itself.
+
+This file aims to sum up how it is supported by Linux and what behaviour you
+can expect from your user programs.
+
+
+Basic overview
+==============
+
+Hardware Transactional Memory is supported on POWER8 processors, and is a
+feature that enables a different form of atomic memory access. Several new
+instructions are presented to delimit transactions; transactions are
+guaranteed to either complete atomically or roll back and undo any partial
+changes.
+
+A simple transaction looks like this:
+
+begin_move_money:
+ tbegin
+ beq abort_handler
+
+ ld r4, SAVINGS_ACCT(r3)
+ ld r5, CURRENT_ACCT(r3)
+ subi r5, r5, 1
+ addi r4, r4, 1
+ std r4, SAVINGS_ACCT(r3)
+ std r5, CURRENT_ACCT(r3)
+
+ tend
+
+ b continue
+
+abort_handler:
+ ... test for odd failures ...
+
+ /* Retry the transaction if it failed because it conflicted with
+ * someone else: */
+ b begin_move_money
+
+
+The 'tbegin' instruction denotes the start point, and 'tend' the end point.
+Between these points the processor is in 'Transactional' state; any memory
+references will complete in one go if there are no conflicts with other
+transactional or non-transactional accesses within the system. In this
+example, the transaction completes as though it were normal straight-line code
+IF no other processor has touched SAVINGS_ACCT(r3) or CURRENT_ACCT(r3); an
+atomic move of money from the current account to the savings account has been
+performed. Even though the normal ld/std instructions are used (note no
+lwarx/stwcx), either *both* SAVINGS_ACCT(r3) and CURRENT_ACCT(r3) will be
+updated, or neither will be updated.
+
+If, in the meantime, there is a conflict with the locations accessed by the
+transaction, the transaction will be aborted by the CPU. Register and memory
+state will roll back to that at the 'tbegin', and control will continue from
+'tbegin+4'. The branch to abort_handler will be taken this second time; the
+abort handler can check the cause of the failure, and retry.
+
+Checkpointed registers include all GPRs, FPRs, VRs/VSRs, LR, CCR/CR, CTR, FPCSR
+and a few other status/flag regs; see the ISA for details.
+
+Causes of transaction aborts
+============================
+
+- Conflicts with cache lines used by other processors
+- Signals
+- Context switches
+- See the ISA for full documentation of everything that will abort transactions.
+
+
+Syscalls
+========
+
+Performing syscalls from within transaction is not recommended, and can lead
+to unpredictable results.
+
+Syscalls do not by design abort transactions, but beware: The kernel code will
+not be running in transactional state. The effect of syscalls will always
+remain visible, but depending on the call they may abort your transaction as a
+side-effect, read soon-to-be-aborted transactional data that should not remain
+invisible, etc. If you constantly retry a transaction that constantly aborts
+itself by calling a syscall, you'll have a livelock & make no progress.
+
+Simple syscalls (e.g. sigprocmask()) "could" be OK. Even things like write()
+from, say, printf() should be OK as long as the kernel does not access any
+memory that was accessed transactionally.
+
+Consider any syscalls that happen to work as debug-only -- not recommended for
+production use. Best to queue them up till after the transaction is over.
+
+
+Signals
+=======
+
+Delivery of signals (both sync and async) during transactions provides a second
+thread state (ucontext/mcontext) to represent the second transactional register
+state. Signal delivery 'treclaim's to capture both register states, so signals
+abort transactions. The usual ucontext_t passed to the signal handler
+represents the checkpointed/original register state; the signal appears to have
+arisen at 'tbegin+4'.
+
+If the sighandler ucontext has uc_link set, a second ucontext has been
+delivered. For future compatibility the MSR.TS field should be checked to
+determine the transactional state -- if so, the second ucontext in uc->uc_link
+represents the active transactional registers at the point of the signal.
+
+For 64-bit processes, uc->uc_mcontext.regs->msr is a full 64-bit MSR and its TS
+field shows the transactional mode.
+
+For 32-bit processes, the mcontext's MSR register is only 32 bits; the top 32
+bits are stored in the MSR of the second ucontext, i.e. in
+uc->uc_link->uc_mcontext.regs->msr. The top word contains the transactional
+state TS.
+
+However, basic signal handlers don't need to be aware of transactions
+and simply returning from the handler will deal with things correctly:
+
+Transaction-aware signal handlers can read the transactional register state
+from the second ucontext. This will be necessary for crash handlers to
+determine, for example, the address of the instruction causing the SIGSEGV.
+
+Example signal handler:
+
+ void crash_handler(int sig, siginfo_t *si, void *uc)
+ {
+ ucontext_t *ucp = uc;
+ ucontext_t *transactional_ucp = ucp->uc_link;
+
+ if (ucp_link) {
+ u64 msr = ucp->uc_mcontext.regs->msr;
+ /* May have transactional ucontext! */
+#ifndef __powerpc64__
+ msr |= ((u64)transactional_ucp->uc_mcontext.regs->msr) << 32;
+#endif
+ if (MSR_TM_ACTIVE(msr)) {
+ /* Yes, we crashed during a transaction. Oops. */
+ fprintf(stderr, "Transaction to be restarted at 0x%llx, but "
+ "crashy instruction was at 0x%llx\n",
+ ucp->uc_mcontext.regs->nip,
+ transactional_ucp->uc_mcontext.regs->nip);
+ }
+ }
+
+ fix_the_problem(ucp->dar);
+ }
+
+
+Failure cause codes used by kernel
+==================================
+
+These are defined in <asm/reg.h>, and distinguish different reasons why the
+kernel aborted a transaction:
+
+ TM_CAUSE_RESCHED Thread was rescheduled.
+ TM_CAUSE_FAC_UNAV FP/VEC/VSX unavailable trap.
+ TM_CAUSE_SYSCALL Currently unused; future syscalls that must abort
+ transactions for consistency will use this.
+ TM_CAUSE_SIGNAL Signal delivered.
+ TM_CAUSE_MISC Currently unused.
+
+These can be checked by the user program's abort handler as TEXASR[0:7].
+
+
+GDB
+===
+
+GDB and ptrace are not currently TM-aware. If one stops during a transaction,
+it looks like the transaction has just started (the checkpointed state is
+presented). The transaction cannot then be continued and will take the failure
+handler route. Furthermore, the transactional 2nd register state will be
+inaccessible. GDB can currently be used on programs using TM, but not sensibly
+in parts within transactions.
--
1.7.10.4
^ permalink raw reply related
* Re: [PATCH v3 1/2] memory-hotplug: introduce CONFIG_HAVE_BOOTMEM_INFO_NODE and revert register_page_bootmem_info_node() when platform not support
From: Lin Feng @ 2013-01-18 7:58 UTC (permalink / raw)
To: Michal Hocko
Cc: linux-ia64, aquini, heiko.carstens, tangchen, dhowells,
paul.gortmaker, paulus, jbeulich, hpa, sparclinux, linux-s390,
x86, isimatu.yasuaki, mingo, gerald.schaefer, fenghua.yu,
jiang.liu, wency, mel, tglx, kamezawa.hiroyu, tony.luck, laijs,
linux-mm, gregkh, linux-kernel, minchan, schwidefsky, linux390,
akpm, linuxppc-dev, davem
In-Reply-To: <20130117130509.GE20538@dhcp22.suse.cz>
Hi Michal,
On 01/17/2013 09:05 PM, Michal Hocko wrote:
> On Thu 17-01-13 18:37:10, Lin Feng wrote:
> [...]
>>> > > I am still not sure I understand the relation to MEMORY_HOTREMOVE.
>>> > > Is register_page_bootmem_info_node required/helpful even if
>>> > > !CONFIG_MEMORY_HOTREMOVE?
>> > From old kenrel's view register_page_bootmem_info_node() is defined in
>> > CONFIG_MEMORY_HOTPLUG_SPARSE, it registers some info for
>> > memory hotplug/remove. If we don't use MEMORY_HOTPLUG feature, this
>> > function is empty, we don't need the info at all.
>> > So this info is not required/helpful if !CONFIG_MEMORY_HOTREMOVE.
> OK, then I suggest moving it under CONFIG_MEMORY_HOTREMOVE guards rather
> than CONFIG_MEMORY_HOTPLUG.
I can't agree more ;-)
I also find that page_isolation.c selected by MEMORY_ISOLATION under MEMORY_HOTPLUG
is also such case, I fix it by the way.
thanks,
linfeng
>
^ permalink raw reply
* [PATCH] powerpc/85xx: dts - add ranges property for SEC
From: Po Liu @ 2013-01-18 9:16 UTC (permalink / raw)
To: linuxppc-dev; +Cc: Po Liu
This facilitates getting the physical address of the SEC node.
Signed-off-by: Liu po <po.liu@freescale.com>
---
arch/powerpc/boot/dts/fsl/pq3-sec4.4-0.dtsi | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/arch/powerpc/boot/dts/fsl/pq3-sec4.4-0.dtsi b/arch/powerpc/boot/dts/fsl/pq3-sec4.4-0.dtsi
index d4c9d5d..ef98fbf 100644
--- a/arch/powerpc/boot/dts/fsl/pq3-sec4.4-0.dtsi
+++ b/arch/powerpc/boot/dts/fsl/pq3-sec4.4-0.dtsi
@@ -36,6 +36,7 @@ crypto@30000 {
compatible = "fsl,sec-v4.4", "fsl,sec-v4.0";
#address-cells = <1>;
#size-cells = <1>;
+ ranges = <0x0 0x30000 0x10000>
reg = <0x30000 0x10000>;
interrupts = <58 2 0 0>;
--
1.6.4
^ permalink raw reply related
* Re: [RFC PATCH 1/6] usb: otg: Add an API to bind the USB controller and PHY
From: Felipe Balbi @ 2013-01-18 11:48 UTC (permalink / raw)
To: kishon
Cc: linux-doc, tony, linux, linux-sh, alexander.shishkin, stern,
devicetree-discuss, linuxppc-dev, rob.herring, horms,
haojian.zhuang, linux-omap, linux-arm-kernel, Roger Quadros,
eric.y.miao, b-cousson, gregkh, linux-usb, linux-kernel, balbi,
cbou, rob, dwmw2
In-Reply-To: <50F7DD2C.90500@ti.com>
[-- Attachment #1: Type: text/plain, Size: 870 bytes --]
On Thu, Jan 17, 2013 at 04:44:52PM +0530, kishon wrote:
> >>@@ -171,6 +188,11 @@ static inline void devm_usb_put_phy(struct device *dev, struct usb_phy *x)
> >> {
> >> }
> >>
> >>+static inline struct usb_phy_bind *usb_bind_phy(const char *dev_name, u8 index,
> >>+ const char *phy_dev_name)
> >>+{
> >>+ return NULL;
> >>+}
> >> #endif
> >>
> >> static inline int
> >>
> >
> >Controllers like ehci-omap which don't need OTG functionality would
> >benefit from this API. Can we make these PHY APIs not dependent on OTG /
> >OTG_UTILS?
>
> Actually much of whatever is in otg.c can be used by controllers
> which don't have OTG functionality (except otg_state_string). I
> vaguely remember, there was a patch that renamed otg.c to phy.c etc..
> I'm not sure what happened to that.
right, that has to be done eventually ;-)
--
balbi
[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]
^ permalink raw reply
* Re: [RFC PATCH 0/6] USB: Add multiple PHYs of same type
From: Felipe Balbi @ 2013-01-18 11:48 UTC (permalink / raw)
To: Kishon Vijay Abraham I
Cc: linux-doc, tony, linux, linux-sh, alexander.shishkin, stern,
devicetree-discuss, linuxppc-dev, rob.herring, horms,
haojian.zhuang, linux-omap, linux-arm-kernel, eric.y.miao,
b-cousson, gregkh, linux-usb, linux-kernel, balbi, cbou, rob,
dwmw2
In-Reply-To: <1358348462-27693-1-git-send-email-kishon@ti.com>
[-- Attachment #1: Type: text/plain, Size: 1201 bytes --]
On Wed, Jan 16, 2013 at 08:30:56PM +0530, Kishon Vijay Abraham I wrote:
> New platforms are being added which has multiple PHY's (of same type) and
> which has multiple USB controllers. The binding information has to be
> present in the PHY library (otg.c) in order for it to return the
> appropriate PHY whenever the USB controller request for the PHY. So
> added a new API to pass the binding information. This API should be
> called by platform specific initialization code.
>
> So the binding should be done something like
> usb_bind_phy("musb-hdrc.0.auto", 0, "omap-usb2.1.auto"); specifying the USB
> controller device name, index, and the PHY device name.
> I have done this binding for OMAP platforms, but it should be done for
> all the platforms.
>
> After this design, the phy can be got by passing the USB controller device
> pointer and the index.
>
> Developed this patch series on
> git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb.git xceiv
> after applying "usb: musb: add driver for control module" patch series.
>
> Did basic enumeration testing in omap4 panda, omap4 sdp and omap3 beagle.
please resend without RFC so I can apply.
--
balbi
[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]
^ permalink raw reply
* Re: [RFC PATCH 0/6] USB: Add multiple PHYs of same type
From: kishon @ 2013-01-18 11:54 UTC (permalink / raw)
To: balbi
Cc: linux-doc, tony, linux, linux-sh, alexander.shishkin, stern,
devicetree-discuss, linuxppc-dev, rob.herring, horms,
haojian.zhuang, linux-omap, linux-arm-kernel, eric.y.miao,
b-cousson, gregkh, linux-usb, linux-kernel, cbou, rob, dwmw2
In-Reply-To: <20130118114859.GC4379@arwen.pp.htv.fi>
On Friday 18 January 2013 05:18 PM, Felipe Balbi wrote:
> On Wed, Jan 16, 2013 at 08:30:56PM +0530, Kishon Vijay Abraham I wrote:
>> New platforms are being added which has multiple PHY's (of same type) and
>> which has multiple USB controllers. The binding information has to be
>> present in the PHY library (otg.c) in order for it to return the
>> appropriate PHY whenever the USB controller request for the PHY. So
>> added a new API to pass the binding information. This API should be
>> called by platform specific initialization code.
>>
>> So the binding should be done something like
>> usb_bind_phy("musb-hdrc.0.auto", 0, "omap-usb2.1.auto"); specifying the USB
>> controller device name, index, and the PHY device name.
>> I have done this binding for OMAP platforms, but it should be done for
>> all the platforms.
>>
>> After this design, the phy can be got by passing the USB controller device
>> pointer and the index.
>>
>> Developed this patch series on
>> git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb.git xceiv
>> after applying "usb: musb: add driver for control module" patch series.
>>
>> Did basic enumeration testing in omap4 panda, omap4 sdp and omap3 beagle.
>
> please resend without RFC so I can apply.
I'll resend after addressing Roger's comments in a while.
Thanks
Kishon
^ permalink raw reply
* Re: [PATCH 6/6][v3] perf: Document the ABI of perf sysfs entries
From: Sukadev Bhattiprolu @ 2013-01-18 17:46 UTC (permalink / raw)
To: Jiri Olsa
Cc: Andi Kleen, Peter Zijlstra, robert.richter, Greg KH,
Anton Blanchard, linux-kernel, Stephane Eranian, linuxppc-dev,
Ingo Molnar, Paul Mackerras, Arnaldo Carvalho de Melo
In-Reply-To: <20130117141145.GA3817@krava.brq.redhat.com>
Jiri Olsa [jolsa@redhat.com] wrote:
| Maybe it'd worth to mention, that it does not need to be just 'event',
| but anything from /sys/bus/event_source/devices/<dev>/format directory,
| like
|
| The can be multiple terms like 'event=0xNNNN' specified
| and separated with comma. All available terms are located
| in /sys/bus/event_source/devices/<dev>/format file.
|
| Please feel free to rephrase or use proper English ;-)
|
| otherwise it's ok,
| jirka
Agree. Thanks for the review. Here is the updated patch.
---
>From 1e3cc6b3ef87f533985b10574af472361e39eecd Mon Sep 17 00:00:00 2001
From: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Date: Tue, 8 Jan 2013 22:31:49 -0800
Subject: [PATCH] perf: Document the ABI of perf sysfs entries
This patchset addes two new sets of files to sysfs for POWER architecture.
- perf event config format in /sys/devices/cpu/format/event
- generic and POWER-specific perf events in /sys/devices/cpu/events/
The format of the first file is already documented in:
sysfs-bus-event_source-devices-format
Document the format of the second set of files '/sys/devices/cpu/events/*'
which would also become part of the ABI.
Changelog[v3.2]: (small changes to this one patch).
[Jiri Olsa]: Mention that multiple event= like terms can be specified
in the 'events' file.
Changelog[v3.1]:
(small changes to just this one patch).
[Jiri Olsa]: Remove the documentation for the 'config format' file
as it is already documented in 'Documentation/ABI/testing/'.
[Jiri Olsa]: Move the documentation of 'events' also to 'testing'
from 'stable'.
Changelog[v3]:
[Greg KH] Include ABI documentation.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
---
.../testing/sysfs-bus-event_source-devices-events | 62 ++++++++++++++++++++
1 files changed, 62 insertions(+), 0 deletions(-)
delete mode 100644 Documentation/ABI/stable/sysfs-devices-cpu-events
create mode 100644 Documentation/ABI/testing/sysfs-bus-event_source-devices-events
diff --git a/Documentation/ABI/stable/sysfs-devices-cpu-events b/Documentation/ABI/stable/sysfs-devices-cpu-events
deleted file mode 100644
index e69de29..0000000
diff --git a/Documentation/ABI/testing/sysfs-bus-event_source-devices-events b/Documentation/ABI/testing/sysfs-bus-event_source-devices-events
new file mode 100644
index 0000000..0adeb52
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-bus-event_source-devices-events
@@ -0,0 +1,62 @@
+What: /sys/devices/cpu/events/
+ /sys/devices/cpu/events/branch-misses
+ /sys/devices/cpu/events/cache-references
+ /sys/devices/cpu/events/cache-misses
+ /sys/devices/cpu/events/stalled-cycles-frontend
+ /sys/devices/cpu/events/branch-instructions
+ /sys/devices/cpu/events/stalled-cycles-backend
+ /sys/devices/cpu/events/instructions
+ /sys/devices/cpu/events/cpu-cycles
+
+Date: 2013/01/08
+
+Contact: Linux kernel mailing list <linux-kernel@vger.kernel.org>
+
+Description: Generic performance monitoring events
+
+ A collection of performance monitoring events that may be
+ supported by many/most CPUs. These events can be monitored
+ using the 'perf(1)' tool.
+
+ The contents of each file would look like:
+
+ event=0xNNNN
+
+ where 'N' is a hex digit and the number '0xNNNN' shows the
+ "raw code" for the perf event identified by the file's
+ "basename".
+
+
+What: /sys/devices/cpu/events/PM_LD_MISS_L1
+ /sys/devices/cpu/events/PM_LD_REF_L1
+ /sys/devices/cpu/events/PM_CYC
+ /sys/devices/cpu/events/PM_BRU_FIN
+ /sys/devices/cpu/events/PM_GCT_NOSLOT_CYC
+ /sys/devices/cpu/events/PM_BRU_MPRED
+ /sys/devices/cpu/events/PM_INST_CMPL
+ /sys/devices/cpu/events/PM_CMPLU_STALL
+
+Date: 2013/01/08
+
+Contact: Linux kernel mailing list <linux-kernel@vger.kernel.org>
+ Linux Powerpc mailing list <linuxppc-dev@ozlabs.org>
+
+Description: POWER-systems specific performance monitoring events
+
+ A collection of performance monitoring events that may be
+ supported by the POWER CPU. These events can be monitored
+ using the 'perf(1)' tool.
+
+ These events may not be supported by other CPUs.
+
+ The contents of each file would look like:
+
+ event=0xNNNN
+
+ where 'N' is a hex digit and the number '0xNNNN' shows the
+ "raw code" for the perf event identified by the file's
+ "basename".
+
+ Further, multiple terms like 'event=0xNNNN' can be specified
+ and separated with comma. All available terms are defined in
+ the /sys/bus/event_source/devices/<dev>/format file.
--
1.7.1
^ permalink raw reply related
* [PATCH] pseries/iommu: Fix iteration in DDW TCE clearrange
From: Nishanth Aravamudan @ 2013-01-18 19:16 UTC (permalink / raw)
To: benh; +Cc: nfont, paulus, linuxppc-dev, miltonm, anton
tce_clearrange_multi_pSeriesLP is attempting to iterate over all TCEs in
a given range. However, is it not advancing the dma_offset value passed
to plpar_tce_stuff via the next value. This prevents DLPAR from
completing, because TCEs are still present at slot isolation time.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
diff --git a/arch/powerpc/platforms/pseries/iommu.c b/arch/powerpc/platforms/pseries/iommu.c
index e2685ba..a55b685 100644
--- a/arch/powerpc/platforms/pseries/iommu.c
+++ b/arch/powerpc/platforms/pseries/iommu.c
@@ -382,6 +382,7 @@ static int tce_clearrange_multi_pSeriesLP(unsigned long start_pfn,
rc = plpar_tce_stuff((u64)be32_to_cpu(maprange->liobn),
dma_offset,
0, limit);
+ next += limit * tce_size;
num_tce -= limit;
} while (num_tce > 0 && !rc);
^ permalink raw reply related
* [PATCH] pseries/iommu: ensure TCEs are cleared with non-huge DDW
From: Nishanth Aravamudan @ 2013-01-18 19:17 UTC (permalink / raw)
To: benh; +Cc: nfont, paulus, linuxppc-dev, miltonm, anton
There are now two kinds of DMA windows that might be presented by
PowerVM DDW support -- huge windows (that can map all of system memory
regardless of the LPAR configuration) and non-huge windows (which
can't). They are implemented slightly differently in PowerVM, and thus
have different characteristics. The most obvious is that slot isolate
doesn't clear the TCEs/window for us with non-huge windows. Thus, when a
DLPAR operation occurs on a slot using a non-huge window, TCEs are still
present (the notifier chain doesn't currently remove them explicitly)
and the DLPAR fails. Fix this by calling remove_ddw() first, which will
unmap the DDW TCEs.
Note: a corresponding change to drmgr is needed to actually successfully
DLPAR, such that the device-tree update (which causes the notifier chain
to fire) occurs before slot isolate.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
diff --git a/arch/powerpc/platforms/pseries/iommu.c b/arch/powerpc/platforms/pseries/iommu.c
index a55b685..b4bb9e1 100644
--- a/arch/powerpc/platforms/pseries/iommu.c
+++ b/arch/powerpc/platforms/pseries/iommu.c
@@ -1296,6 +1296,7 @@ static int iommu_reconfig_notifier(struct notifier_block *nb, unsigned long acti
switch (action) {
case OF_RECONFIG_DETACH_NODE:
+ remove_ddw(np);
if (pci && pci->iommu_table)
iommu_free_table(pci->iommu_table, np->full_name);
@@ -1308,16 +1309,6 @@ static int iommu_reconfig_notifier(struct notifier_block *nb, unsigned long acti
}
}
spin_unlock(&direct_window_list_lock);
-
- /*
- * Because the notifier runs after isolation of the
- * slot, we are guaranteed any DMA window has already
- * been revoked and the TCEs have been marked invalid,
- * so we don't need a call to remove_ddw(np). However,
- * if an additional notifier action is added before the
- * isolate call, we should update this code for
- * completeness with such a call.
- */
break;
default:
err = NOTIFY_DONE;
^ permalink raw reply related
* Re: [PATCH 6/6][v3] perf: Document the ABI of perf sysfs entries
From: Arnaldo Carvalho de Melo @ 2013-01-18 19:20 UTC (permalink / raw)
To: Sukadev Bhattiprolu
Cc: Andi Kleen, Peter Zijlstra, robert.richter, Greg KH,
Anton Blanchard, linux-kernel, Stephane Eranian, linuxppc-dev,
Ingo Molnar, Paul Mackerras, Jiri Olsa
In-Reply-To: <20130118174654.GA12575@us.ibm.com>
Em Fri, Jan 18, 2013 at 09:46:54AM -0800, Sukadev Bhattiprolu escreveu:
> Jiri Olsa [jolsa@redhat.com] wrote:
> | Maybe it'd worth to mention, that it does not need to be just 'event',
> | but anything from /sys/bus/event_source/devices/<dev>/format directory,
> | like
> |
> | The can be multiple terms like 'event=0xNNNN' specified
> | and separated with comma. All available terms are located
> | in /sys/bus/event_source/devices/<dev>/format file.
> |
> | Please feel free to rephrase or use proper English ;-)
> |
> | otherwise it's ok,
> | jirka
>
> Agree. Thanks for the review. Here is the updated patch.
Ok, I'll wait for Jiri's final acked-by, then you can resubmit the patch
series and I'll apply it to my perf/core branch.
- Arnaldo
^ permalink raw reply
* Re: [PATCH] powerpc/85xx: dts - add ranges property for SEC
From: Kim Phillips @ 2013-01-18 20:40 UTC (permalink / raw)
To: Po Liu; +Cc: linuxppc-dev
In-Reply-To: <1358500573-19477-1-git-send-email-po.liu@freescale.com>
On Fri, 18 Jan 2013 17:16:13 +0800
Po Liu <po.liu@freescale.com> wrote:
> This facilitates getting the physical address of the SEC node.
>
> Signed-off-by: Liu po <po.liu@freescale.com>
> ---
Reviewed-by: Kim Phillips <kim.phillips@freescale.com>
Kim
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox