Skip to contentSkip to conversation

hw_breakpoint fixes

GIT PULL New Frederic Weisbecker 2 participants 4 messages Sep 17, 2010 x86

lore
Share & tools
lore
#0 lore raw

Ingo,

Please pull the perf/urgent branch that can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git
perf/urgent

Thanks,
Frederic
---

Frederic Weisbecker (1):
x86: Fix instruction breakpoint encoding

Matt Helsley (1):
hw breakpoints: Fix pid namespace bug

 arch/x86/include/asm/hw_breakpoint.h |    2 +-
 arch/x86/kernel/hw_breakpoint.c      |   40 ++++++++++++++++-----------------
 kernel/hw_breakpoint.c               |    3 +-
 3 files changed, 22 insertions(+), 23 deletions(-)
#1 lore raw

PATCH 1/2 x86: Fix instruction breakpoint encoding

Lengths and types of breakpoints are encoded in a half byte into CPU registers. However when we extract these values and store them, we add a high half byte part to them: 0x40 to the length and 0x80 to the type.
When that gets reloaded to the CPU registers, the high part is masked.

While making the instruction breakpoints available for perf, I zapped that high part on instruction breakpoint encoding and that broke the arch -> generic translation used by ptrace instruction breakpoints. Writing dr7 to set an inst breakpoint was then failing.

There is no apparent reason for these high parts so we could get rid of them altogether. That's an invasive change though so let's do that later and for now fix the problem by restoring that inst breakpoint high part encoding in this sole patch.

---
 arch/x86/include/asm/hw_breakpoint.h |    2 +-
 arch/x86/kernel/hw_breakpoint.c      |   40 ++++++++++++++++-----------------
 2 files changed, 20 insertions(+), 22 deletions(-)
arch/x86/include/asm/hw_breakpoint.h +1 -1
index 528a11e..824ca07 100644
@@ -20,7 +20,7 @@ struct arch_hw_breakpoint {
 #include <linux/list.h>
 
 /* Available HW breakpoint length encodings */
-#define X86_BREAKPOINT_LEN_X		0x00
+#define X86_BREAKPOINT_LEN_X		0x40
 #define X86_BREAKPOINT_LEN_1		0x40
 #define X86_BREAKPOINT_LEN_2		0x44
 #define X86_BREAKPOINT_LEN_4		0x4c
arch/x86/kernel/hw_breakpoint.c +19 -21
index a474ec3..ff15c9d 100644
@@ -206,11 +206,27 @@ int arch_check_bp_in_kernelspace(struct perf_event *bp)
 int arch_bp_generic_fields(int x86_len, int x86_type,
 			   int *gen_len, int *gen_type)
 {
-	/* Len */
-	switch (x86_len) {
-	case X86_BREAKPOINT_LEN_X:
+	/* Type */
+	switch (x86_type) {
+	case X86_BREAKPOINT_EXECUTE:
+		if (x86_len != X86_BREAKPOINT_LEN_X)
+			return -EINVAL;
+
+		*gen_type = HW_BREAKPOINT_X;
 		*gen_len = sizeof(long);
+		return 0;
+	case X86_BREAKPOINT_WRITE:
+		*gen_type = HW_BREAKPOINT_W;
 		break;
+	case X86_BREAKPOINT_RW:
+		*gen_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Len */
+	switch (x86_len) {
 	case X86_BREAKPOINT_LEN_1:
 		*gen_len = HW_BREAKPOINT_LEN_1;
 		break;
@@ -229,21 +245,6 @@ int arch_bp_generic_fields(int x86_len, int x86_type,
 		return -EINVAL;
 	}
 
-	/* Type */
-	switch (x86_type) {
-	case X86_BREAKPOINT_EXECUTE:
-		*gen_type = HW_BREAKPOINT_X;
-		break;
-	case X86_BREAKPOINT_WRITE:
-		*gen_type = HW_BREAKPOINT_W;
-		break;
-	case X86_BREAKPOINT_RW:
-		*gen_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R;
-		break;
-	default:
-		return -EINVAL;
-	}
-
 	return 0;
 }
 
@@ -316,9 +317,6 @@ int arch_validate_hwbkpt_settings(struct perf_event *bp)
 	ret = -EINVAL;
 
 	switch (info->len) {
-	case X86_BREAKPOINT_LEN_X:
-		align = sizeof(long) -1;
-		break;
 	case X86_BREAKPOINT_LEN_1:
 		align = 0;
 		break;
Signature · 1 line
1.6.2.3
#2 lore raw

PATCH 2/2 hw breakpoints: Fix pid namespace bug

In reply to Frederic Weisbecker

From: Matt Helsley <[email protected]>

Hardware breakpoints can't be registered within pid namespaces because tsk->pid is passed rather than the pid in the current namespace.

(See https://bugzilla.kernel.org/show_bug.cgi?id=17281 )

This is a quick fix demonstrating the problem but is not the best method of solving the problem since passing pids internally is not the best way to avoid pid namespace bugs. Subsequent patches will show a better solution.

Much thanks to Frederic Weisbecker <[email protected]> for doing the bulk of the work finding this bug.

Cc Prasad <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Steven Rostedt <[email protected]>
Cc Will Deacon <[email protected]>
Cc Mahesh Salgaonkar <[email protected]>
Cc 2.6.33-2.6.35 <[email protected]>

LKML-Reference: <f63454af09fb1915717251570423eb9ddd338340.1284407762.git.matthltc@us.ibm.com>

---
 kernel/hw_breakpoint.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
kernel/hw_breakpoint.c +2 -1
index d71a987..c7c2aed 100644
@@ -433,7 +433,8 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 			    perf_overflow_handler_t triggered,
 			    struct task_struct *tsk)
 {
-	return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered);
+	return perf_event_create_kernel_counter(attr, -1, task_pid_vnr(tsk),
+						triggered);
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
Signature · 1 line
1.6.2.3
#3 lore raw
In reply to Frederic Weisbecker
Quoted 14 lines · Frederic Weisbecker

Ingo,

Please pull the perf/urgent branch that can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git
perf/urgent

Thanks,
Frederic
---

Frederic Weisbecker (1):
x86: Fix instruction breakpoint encoding

Quoted 8 lines

Matt Helsley (1):
hw breakpoints: Fix pid namespace bug

 arch/x86/include/asm/hw_breakpoint.h |    2 +-
 arch/x86/kernel/hw_breakpoint.c      |   40 ++++++++++++++++-----------------
 kernel/hw_breakpoint.c               |    3 +-
 3 files changed, 22 insertions(+), 23 deletions(-)

Pulled, thanks a lot Frederic!

Ingo