[PATCH] powerpc/perf: Dereference bhrb entries safely

It may very well happen that branch instructions recorded by
bhrb entries already get unmapped before they get processed by
the kernel. Hence, trying to dereference such memory location
will endup in a crash. Ex,

    Unable to handle kernel paging request for data at address 0xc008000019c41764
    Faulting instruction address: 0xc000000000084a14
    NIP [c000000000084a14] branch_target+0x4/0x70
    LR [c0000000000eb828] record_and_restart+0x568/0x5c0
    Call Trace:
    [c0000000000eb3b4] record_and_restart+0xf4/0x5c0 (unreliable)
    [c0000000000ec378] perf_event_interrupt+0x298/0x460
    [c000000000027964] performance_monitor_exception+0x54/0x70
    [c000000000009ba4] performance_monitor_common+0x114/0x120

Fix this by deferefencing them safely.

Suggested-by: Naveen N. Rao <redacted>
Signed-off-by: Ravi Bangoria <redacted>
---
 arch/powerpc/perf/core-book3s.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c
index 9e3da168d54c..5a68d2effdf9 100644
--- a/arch/powerpc/perf/core-book3s.c
+++ b/arch/powerpc/perf/core-book3s.c

@@ -410,8 +410,11 @@ static __u64 power_pmu_bhrb_to(u64 addr)
 	int ret;
 	__u64 target;
 
-	if (is_kernel_addr(addr))
-		return branch_target((unsigned int *)addr);
+	if (is_kernel_addr(addr)) {
+		if (probe_kernel_address((void *)addr, instr))
+			return 0;
+		return branch_target(&instr);
+	}
 
 	/* Userspace: need copy instruction here then translate it */
 	pagefault_disable();

-- 
2.13.6

Ravi Bangoria wrote:

It may very well happen that branch instructions recorded by
bhrb entries already get unmapped before they get processed by
the kernel. Hence, trying to dereference such memory location
will endup in a crash. Ex,
=20
    Unable to handle kernel paging request for data at address 0xc0080000=

19c41764

    Faulting instruction address: 0xc000000000084a14
    NIP [c000000000084a14] branch_target+0x4/0x70
    LR [c0000000000eb828] record_and_restart+0x568/0x5c0
    Call Trace:
    [c0000000000eb3b4] record_and_restart+0xf4/0x5c0 (unreliable)
    [c0000000000ec378] perf_event_interrupt+0x298/0x460
    [c000000000027964] performance_monitor_exception+0x54/0x70
    [c000000000009ba4] performance_monitor_common+0x114/0x120
=20
Fix this by deferefencing them safely.
=20
Suggested-by: Naveen N. Rao <redacted>
Signed-off-by: Ravi Bangoria <redacted>

Reviewed-by: Naveen N. Rao <redacted>


- Naveen

=

On Tue, Dec 12, 2017 at 11:29 PM, Ravi Bangoria
[off-list ref] wrote:

I think __kernel_text_address() is more accurate right? In which case
you need to check for is_kernel_addr(addr) and if its not kernel_text_address()
then we have an interesting case of a branch from something not text.
It would be nice to catch such cases.

+               if (probe_kernel_address((void *)addr, instr))
+                       return 0;
+               return branch_target(&instr);
+       }

        /* Userspace: need copy instruction here then translate it */
        pagefault_disable();

Otherwise,

Reviewed-by: Balbir Singh <bsingharora@gmail.com>

Hi Balbir,

Sorry was away for few days.

On 12/14/2017 05:54 PM, Balbir Singh wrote:

On Tue, Dec 12, 2017 at 11:29 PM, Ravi Bangoria
[off-list ref] wrote:

quoted

It may very well happen that branch instructions recorded by
bhrb entries already get unmapped before they get processed by
the kernel. Hence, trying to dereference such memory location
will endup in a crash. Ex,

    Unable to handle kernel paging request for data at address 0xc008000019c41764
    Faulting instruction address: 0xc000000000084a14
    NIP [c000000000084a14] branch_target+0x4/0x70
    LR [c0000000000eb828] record_and_restart+0x568/0x5c0
    Call Trace:
    [c0000000000eb3b4] record_and_restart+0xf4/0x5c0 (unreliable)
    [c0000000000ec378] perf_event_interrupt+0x298/0x460
    [c000000000027964] performance_monitor_exception+0x54/0x70
    [c000000000009ba4] performance_monitor_common+0x114/0x120

Fix this by deferefencing them safely.

Suggested-by: Naveen N. Rao <redacted>
Signed-off-by: Ravi Bangoria <redacted>
---
 arch/powerpc/perf/core-book3s.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c
index 9e3da168d54c..5a68d2effdf9 100644
--- a/arch/powerpc/perf/core-book3s.c
+++ b/arch/powerpc/perf/core-book3s.c

@@ -410,8 +410,11 @@ static __u64 power_pmu_bhrb_to(u64 addr)
        int ret;
        __u64 target;

-       if (is_kernel_addr(addr))
-               return branch_target((unsigned int *)addr);
+       if (is_kernel_addr(addr)) {

I think __kernel_text_address() is more accurate right? In which case
you need to check for is_kernel_addr(addr) and if its not kernel_text_address()
then we have an interesting case of a branch from something not text.
It would be nice to catch such cases.

Something like this?

diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c
index 1538129..627af56 100644
--- a/arch/powerpc/perf/core-book3s.c
+++ b/arch/powerpc/perf/core-book3s.c

@@ -410,8 +410,13 @@ static __u64 power_pmu_bhrb_to(u64 addr)

     int ret;
     __u64 target;
 
-    if (is_kernel_addr(addr))
-        return branch_target((unsigned int *)addr);
+    if (is_kernel_addr(addr)) {
+        if (probe_kernel_address((void *)addr, instr)) {
+            WARN_ON(!__kernel_text_address(addr));
+            return 0;
+        }
+        return branch_target(&instr);
+    }
 
     /* Userspace: need copy instruction here then translate it */
     pagefault_disable();


I think this will throw warnings when you try to read recently unmapped
vmalloced address. Is that fine?

Thanks for the review.
Ravi

quoted

+               if (probe_kernel_address((void *)addr, instr))
+                       return 0;
+               return branch_target(&instr);
+       }

        /* Userspace: need copy instruction here then translate it */
        pagefault_disable();

Otherwise,

Reviewed-by: Balbir Singh <bsingharora@gmail.com>

On Tue, Dec 19, 2017 at 6:23 PM, Ravi Bangoria
[off-list ref] wrote:

Hi Balbir,

Sorry was away for few days.

No problem at all

More like if (__kernel_text_address(addr))
                 if (probe_kernel_address(...))

+        if (probe_kernel_address((void *)addr, instr)) {
+            WARN_ON(!__kernel_text_address(addr));
+            return 0;
+        }
+        return branch_target(&instr);
+    }

     /* Userspace: need copy instruction here then translate it */
     pagefault_disable();


I think this will throw warnings when you try to read recently unmapped
vmalloced address. Is that fine?

I'd rather we not probe addresses that are not text for this case. if
it is unmapped
that is a challenge, but that might happen for unloaded modules and eBPF code
(hopefully that will be rare)

Balbir Singh

Ravi Bangoria [off-list ref] writes:

Hi Balbir,

Sorry was away for few days.

On 12/14/2017 05:54 PM, Balbir Singh wrote:

quoted

On Tue, Dec 12, 2017 at 11:29 PM, Ravi Bangoria
[off-list ref] wrote:

quoted

It may very well happen that branch instructions recorded by
bhrb entries already get unmapped before they get processed by
the kernel. Hence, trying to dereference such memory location
will endup in a crash. Ex,

    Unable to handle kernel paging request for data at address 0xc00800=

0019c41764

quoted

quoted

    Faulting instruction address: 0xc000000000084a14
    NIP [c000000000084a14] branch_target+0x4/0x70
    LR [c0000000000eb828] record_and_restart+0x568/0x5c0
    Call Trace:
    [c0000000000eb3b4] record_and_restart+0xf4/0x5c0 (unreliable)
    [c0000000000ec378] perf_event_interrupt+0x298/0x460
    [c000000000027964] performance_monitor_exception+0x54/0x70
    [c000000000009ba4] performance_monitor_common+0x114/0x120

Fix this by deferefencing them safely.

Suggested-by: Naveen N. Rao <redacted>
Signed-off-by: Ravi Bangoria <redacted>
---
 arch/powerpc/perf/core-book3s.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-b=

ook3s.c

quoted

quoted

index 9e3da168d54c..5a68d2effdf9 100644

--- a/arch/powerpc/perf/core-book3s.c
+++ b/arch/powerpc/perf/core-book3s.c

@@ -410,8 +410,11 @@ static __u64 power_pmu_bhrb_to(u64 addr)
        int ret;
        __u64 target;

-       if (is_kernel_addr(addr))
-               return branch_target((unsigned int *)addr);
+       if (is_kernel_addr(addr)) {

I think __kernel_text_address() is more accurate right? In which case
you need to check for is_kernel_addr(addr) and if its not kernel_text_ad=

dress()

k3s.c

 *)addr);

+=C2=A0=C2=A0=C2=A0 if (is_kernel_addr(addr)) {
+=C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 if (probe_kernel_address((void *)a=

ddr, instr)) {

+=C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 WARN_ON(!__kern=

el_text_address(addr));

+=C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 return 0;
+=C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 }
+=C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 return branch_target(&instr);
+=C2=A0=C2=A0=C2=A0 }
=C2=A0
=C2=A0=C2=A0=C2=A0=C2=A0 /* Userspace: need copy instruction here then tr=

anslate it */

=C2=A0=C2=A0=C2=A0=C2=A0 pagefault_disable();


I think this will throw warnings when you try to read recently unmapped
vmalloced address. Is that fine?

No.

I've already merged the patch, and am fairly happy with it.

cheers

On Tue, 2017-12-12 at 12:29:15 UTC, Ravi Bangoria wrote:

It may very well happen that branch instructions recorded by
bhrb entries already get unmapped before they get processed by
the kernel. Hence, trying to dereference such memory location
will endup in a crash. Ex,

    Unable to handle kernel paging request for data at address 0xc008000019c41764
    Faulting instruction address: 0xc000000000084a14
    NIP [c000000000084a14] branch_target+0x4/0x70
    LR [c0000000000eb828] record_and_restart+0x568/0x5c0
    Call Trace:
    [c0000000000eb3b4] record_and_restart+0xf4/0x5c0 (unreliable)
    [c0000000000ec378] perf_event_interrupt+0x298/0x460
    [c000000000027964] performance_monitor_exception+0x54/0x70
    [c000000000009ba4] performance_monitor_common+0x114/0x120

Fix this by deferefencing them safely.

Suggested-by: Naveen N. Rao <redacted>
Signed-off-by: Ravi Bangoria <redacted>
Reviewed-by: Naveen N. Rao <redacted>

Applied to powerpc fixes, thanks.

https://git.kernel.org/powerpc/c/f41d84dddc66b164ac16acf3f584c2

cheers