Skip to contentSkip to conversation

arm64, makedumpfile: A couple of KASLR related fixes

PATCH In review Bhupesh Sharma 2 participants 10 messages Jul 19, 2018 – Aug 20, 2018

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover arm64, makedumpfile: A couple of KASLR related fixes

This patchset contains a couple of patches which provide important arm64 fixes for KASLR boot cases:

PATCH 1 -> Addresses the issue with 'info->page_offset' calculation for
KASLR cases.
PATCH 2 -> Addresses the issue with filtering not able to translate
symbol address from vmlinux to kernel run time address in case of KASLR boot.

Bhupesh Sharma (2):
arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR
boot cases
arm64: Add runtime kaslr offset if it exists

 arch/arm64.c   | 64 +++++++++++++++++++++++++++++++++++++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  4 +++-
 3 files changed, 63 insertions(+), 6 deletions(-)
Signature · 7 lines
2.7.4


_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
#1 lore raw

PATCH 1/2 arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

The existing methodology to obtain 'info->page_offset' from reading _stext symbol (from kallsyms) doesn't work well in KASLR boot cases on arm64 machines as the PAGE_OFFSET (or the virtual address which indicates the start of the linear region) can be randomized as well on basis of the kaslr-seed.

Since the value of PAGE_OFFSET inside the kernel is randomized in such cases and there is no existing mechanism of conveying this value from kernel-space to user-space, so we can use the method used by archs like x86_64 to generate the 'info->page_offset' value from the PT_LOAD segments by subtracting the phy_addr from virt_addr of a PT_LOAD segment.

This approach works fine both with KASLR and non-KASLR boot cases.

I tested this on my qualcomm-amberwing board. Here are some logs from the KASLR boot cases:

- Verify that the EFI firmware supports 'kaslr-seed':

	chosen {
		kaslr-seed = <0x0 0x0>;
		<..snip..>
	};

- Verify that '--mem-usage' works well after this fix as well (I used kernel 4.18.0-rc4+ for my checks):

The kernel version is not supported.
The makedumpfile operation may be incomplete.

TYPE PAGES EXCLUDABLE DESCRIPTION
----------------------------------------------------------------------
ZERO 4396 yes Pages filled
with zero
NON_PRI_CACHE 27859 yes Cache pages
without private flag
PRI_CACHE 18490 yes Cache pages with
private flag
USER 2728 yes User process
pages
FREE 1465848 yes Free pages
KERN_DATA 18537 no Dumpable kernel
data

page size: 65536
Total pages on system: 1537858
Total size on system: 100785061888 Byte

---
 arch/arm64.c   | 23 ++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  1 +
 3 files changed, 20 insertions(+), 5 deletions(-)
arch/arm64.c +18 -5
index 2fd3e1874376..9e8c77c76935 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+	int i;
+	unsigned long long phys_start;
+	unsigned long long virt_start;
 	ulong _stext;
 
 	_stext = get_stext_symbol();
@@ -289,12 +292,22 @@ get_versiondep_info_arm64(void)
 		return FALSE;
 	}
 
-	info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
-
-	DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
-			va_bits);
+	if (get_num_pt_loads()) {
+		for (i = 0;
+			get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+			i++) {
+			if (virt_start != NOT_KV_ADDR
+					&& virt_start < __START_KERNEL_map
+					&& phys_start != NOT_PADDR && phys_start != NOT_PADDR_ARM64) {
+				info->page_offset = virt_start - phys_start;
+				DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+						info->page_offset, va_bits);
+				return TRUE;
+			}
+		}
+	}
 
-	return TRUE;
+	return FALSE;
 }
 
 /*
common.h +1 -0
index 6e2f657a79c7..a8181777dbb7 100644
@@ -48,6 +48,7 @@
 #define NOT_MEMMAP_ADDR	(0x0)
 #define NOT_KV_ADDR	(0x0)
 #define NOT_PADDR	(ULONGLONG_MAX)
+#define NOT_PADDR_ARM64	(0x0000000010a80000UL)
 #define BADADDR  	((ulong)(-1))
 
 #endif  /* COMMON_H */
makedumpfile.h +1 -0
index 5ff94b8e4ac6..5297279f0f3b 100644
@@ -2020,6 +2020,7 @@ struct domain_list {
 #define MFNS_PER_FRAME		(info->page_size / sizeof(unsigned long))
 
 #ifdef __aarch64__
+#define __START_KERNEL_map		(0xffffffff80000000UL)
 unsigned long long kvtop_xen_arm64(unsigned long kvaddr);
 #define kvtop_xen(X)	kvtop_xen_arm64(X)
 #endif /* aarch64 */
Signature · 7 lines
2.7.4


_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
#2 lore raw

Re: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Quoted 71 lines · Bhupesh Sharma

The existing methodology to obtain 'info->page_offset' from reading _stext symbol (from kallsyms) doesn't work well in KASLR boot cases on arm64 machines as the PAGE_OFFSET (or the virtual address which indicates the start of the linear region) can be randomized as well on basis of the kaslr-seed.

Since the value of PAGE_OFFSET inside the kernel is randomized in such cases and there is no existing mechanism of conveying this value from kernel-space to user-space, so we can use the method used by archs like x86_64 to generate the 'info->page_offset' value from the PT_LOAD segments by subtracting the phy_addr from virt_addr of a PT_LOAD segment.

This approach works fine both with KASLR and non-KASLR boot cases.

I tested this on my qualcomm-amberwing board. Here are some logs from the KASLR boot cases:

- Verify that the EFI firmware supports 'kaslr-seed':

	chosen {
		kaslr-seed = <0x0 0x0>;
		<..snip..>
	};

- Verify that '--mem-usage' works well after this fix as well (I used kernel 4.18.0-rc4+ for my checks):

The kernel version is not supported.
The makedumpfile operation may be incomplete.

TYPE PAGES EXCLUDABLE DESCRIPTION
----------------------------------------------------------------------
ZERO 4396 yes Pages filled
with zero
NON_PRI_CACHE 27859 yes Cache pages
without private flag
PRI_CACHE 18490 yes Cache pages with
private flag
USER 2728 yes User process
pages
FREE 1465848 yes Free pages
KERN_DATA 18537 no Dumpable kernel
data

page size: 65536
Total pages on system: 1537858
Total size on system: 100785061888 Byte

---
 arch/arm64.c   | 23 ++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  1 +
 3 files changed, 20 insertions(+), 5 deletions(-)
arch/arm64.c +3 -0
index 2fd3e1874376..9e8c77c76935 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+	int i;
+	unsigned long long phys_start;
+	unsigned long long virt_start;
 	ulong _stext;
 
 	_stext = get_stext_symbol();

arch/arm64.c get_versiondep_info_arm64(void) · line 296

Quoted 8 lines
arch/arm64.c +0 -4
@@ -289,12 +292,22 @@ get_versiondep_info_arm64(void)
 		return FALSE;
 	}
 
 
-	info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
-
-	DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
-			va_bits);

According to makedumpfile commit 4944f93484 ("x86_64: Calculate page_offset in case of re-filtering/sadump/virsh dump"), in case of re-filtering, we don't have any PT_LOAD. So, with this patch, we cannot re-filter dump file on arm64 system?

If we cannot, is it better to leave the above behind the following pt_load section for re-filtering non-KASLR dump?

Quoted 4 lines
arch/arm64.c +4 -0
+	if (get_num_pt_loads()) {
+		for (i = 0;
+			get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+			i++) {

arch/arm64.c get_versiondep_info_arm64(void) · line 309

Quoted 10 lines
arch/arm64.c +10 -0
+			if (virt_start != NOT_KV_ADDR
+					&& virt_start < __START_KERNEL_map
+					&& phys_start != NOT_PADDR && phys_start != NOT_PADDR_ARM64) {
+				info->page_offset = virt_start - phys_start;
+				DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+						info->page_offset, va_bits);
+				return TRUE;
+			}
+		}
+	}

I'll adjust some indents for readability.

Quoted 11 lines
arch/arm64.c +1 -1
-	return TRUE;
+	return FALSE;
 }
 
 /*
common.h +0 -0
index 6e2f657a79c7..a8181777dbb7 100644

common.h line 52

Quoted 5 lines
common.h +1 -0
@@ -48,6 +48,7 @@
 #define NOT_MEMMAP_ADDR	(0x0)
 #define NOT_KV_ADDR	(0x0)
 #define NOT_PADDR	(ULONGLONG_MAX)
 
+#define NOT_PADDR_ARM64	(0x0000000010a80000UL)

I think that this should not be in common.h, because it's not for arch-specific definitions, so I'll move it to makedumpfile.h.

makedumpfile.h struct domain_list { · line 2023

Quoted 12 lines
common.h +0 -0
 #define BADADDR  	((ulong)(-1))
 

#endif /* COMMON_H */

makedumpfile.h +1 -0
index 5ff94b8e4ac6..5297279f0f3b 100644
@@ -2020,6 +2020,7 @@ struct domain_list {
 #define MFNS_PER_FRAME		(info->page_size / sizeof(unsigned long))
 
 #ifdef __aarch64__
+#define __START_KERNEL_map		(0xffffffff80000000UL)

This #ifdef here is for the definitions related to Xen extraction. I'll move it to the same place as the above.

makedumpfile.h struct domain_list { · line 2026

Quoted 4 lines
makedumpfile.h +0 -0
 unsigned long long kvtop_xen_arm64(unsigned long kvaddr);
 #define kvtop_xen(X)	kvtop_xen_arm64(X)
 #endif /* aarch64 */

I attached a patch that was applied my suggestions. How about it?

Thanks,
Kazu

arch/arm64.c +19 -1
index 2fd3e18..836ce17 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+	int i;
+	unsigned long long phys_start;
+	unsigned long long virt_start;
 	ulong _stext;
 
 	_stext = get_stext_symbol();
@@ -289,8 +292,23 @@ get_versiondep_info_arm64(void)
 		return FALSE;
 	}
 
-	info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
+	if (get_num_pt_loads()) {
+		for (i = 0;
+		    get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+		    i++) {
+			if (virt_start != NOT_KV_ADDR
+			    && virt_start < __START_KERNEL_map
+			    && phys_start != NOT_PADDR
+			    && phys_start != NOT_PADDR_ARM64) {
+				info->page_offset = virt_start - phys_start;
+				DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+						info->page_offset, va_bits);
+				return TRUE;
+			}
+		}
+	}
 
+	info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
 	DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
 			va_bits);
 
makedumpfile.h +4 -0
index 5ff94b8..ba85f03 100644
@@ -542,6 +542,10 @@ do { \
 #ifdef __aarch64__
 unsigned long get_kvbase_arm64(void);
 #define KVBASE			get_kvbase_arm64()
+
+#define __START_KERNEL_map	(0xffffffff80000000UL)
+#define NOT_PADDR_ARM64		(0x0000000010a80000UL)
+
 #endif /* aarch64 */
 
 #ifdef __arm__
Signature · 1 line
1.8.3.1

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#3 lore raw

Re: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Hello Kazu,

Many thanks for your review comments.

Quoted 143 lines · Kazuhito Hagio
Quoted 71 lines · Bhupesh Sharma

The existing methodology to obtain 'info->page_offset' from reading _stext symbol (from kallsyms) doesn't work well in KASLR boot cases on arm64 machines as the PAGE_OFFSET (or the virtual address which indicates the start of the linear region) can be randomized as well on basis of the kaslr-seed.

Since the value of PAGE_OFFSET inside the kernel is randomized in such cases and there is no existing mechanism of conveying this value from kernel-space to user-space, so we can use the method used by archs like x86_64 to generate the 'info->page_offset' value from the PT_LOAD segments by subtracting the phy_addr from virt_addr of a PT_LOAD segment.

This approach works fine both with KASLR and non-KASLR boot cases.

I tested this on my qualcomm-amberwing board. Here are some logs from the KASLR boot cases:

- Verify that the EFI firmware supports 'kaslr-seed':

      chosen {
              kaslr-seed = <0x0 0x0>;
              <..snip..>
      };

- Verify that '--mem-usage' works well after this fix as well (I used kernel 4.18.0-rc4+ for my checks):

The kernel version is not supported.
The makedumpfile operation may be incomplete.

TYPE PAGES EXCLUDABLE DESCRIPTION
----------------------------------------------------------------------
ZERO 4396 yes Pages filled
with zero
NON_PRI_CACHE 27859 yes Cache pages
without private flag
PRI_CACHE 18490 yes Cache pages with
private flag
USER 2728 yes User process
pages

FREE          1465848                 yes             Free pages
KERN_DATA     18537                   no              Dumpable kernel

data

page size:            65536
Total pages on system:        1537858
Total size on system: 100785061888     Byte
---
 arch/arm64.c   | 23 ++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  1 +
 3 files changed, 20 insertions(+), 5 deletions(-)
arch/arm64.c +3 -0
index 2fd3e1874376..9e8c77c76935 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+     int i;
+     unsigned long long phys_start;
+     unsigned long long virt_start;
      ulong _stext;
 
      _stext = get_stext_symbol();

arch/arm64.c get_versiondep_info_arm64(void) · line 296

Quoted 8 lines
arch/arm64.c +0 -4
@@ -289,12 +292,22 @@ get_versiondep_info_arm64(void)
              return FALSE;
      }
 
 
-     info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
-
-     DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
-                     va_bits);

According to makedumpfile commit 4944f93484 ("x86_64: Calculate page_offset in case of re-filtering/sadump/virsh dump"), in case of re-filtering, we don't have any PT_LOAD. So, with this patch, we cannot re-filter dump file on arm64 system?

If we cannot, is it better to leave the above behind the following pt_load section for re-filtering non-KASLR dump?

Quoted 4 lines
arch/arm64.c +4 -0
+     if (get_num_pt_loads()) {
+             for (i = 0;
+                     get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+                     i++) {

arch/arm64.c get_versiondep_info_arm64(void) · line 309

Quoted 10 lines
arch/arm64.c +10 -0
+                     if (virt_start != NOT_KV_ADDR
+                                     && virt_start < __START_KERNEL_map
+                                     && phys_start != NOT_PADDR && phys_start != NOT_PADDR_ARM64) {
+                             info->page_offset = virt_start - phys_start;
+                             DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+                                             info->page_offset, va_bits);
+                             return TRUE;
+                     }
+             }
+     }

I'll adjust some indents for readability.

Quoted 11 lines
arch/arm64.c +1 -1
-     return TRUE;
+     return FALSE;
 }
 
 /*
common.h +0 -0
index 6e2f657a79c7..a8181777dbb7 100644

common.h line 52

Quoted 5 lines
common.h +1 -0
@@ -48,6 +48,7 @@
 #define NOT_MEMMAP_ADDR      (0x0)
 #define NOT_KV_ADDR  (0x0)
 #define NOT_PADDR    (ULONGLONG_MAX)
 
+#define NOT_PADDR_ARM64      (0x0000000010a80000UL)

I think that this should not be in common.h, because it's not for arch-specific definitions, so I'll move it to makedumpfile.h.

makedumpfile.h struct domain_list { · line 2023

Quoted 12 lines
common.h +0 -0
 #define BADADDR      ((ulong)(-1))
 

#endif /* COMMON_H */

makedumpfile.h +1 -0
index 5ff94b8e4ac6..5297279f0f3b 100644
@@ -2020,6 +2020,7 @@ struct domain_list {
 #define MFNS_PER_FRAME               (info->page_size / sizeof(unsigned long))
 
 #ifdef __aarch64__
+#define __START_KERNEL_map           (0xffffffff80000000UL)

This #ifdef here is for the definitions related to Xen extraction. I'll move it to the same place as the above.

Quoted 7 lines

makedumpfile.h struct domain_list { · line 2026

Quoted 4 lines
makedumpfile.h +0 -0
 unsigned long long kvtop_xen_arm64(unsigned long kvaddr);
 #define kvtop_xen(X) kvtop_xen_arm64(X)
 #endif /* aarch64 */

I attached a patch that was applied my suggestions. How about it?

I agree. The patch looks good with the suggested changes. Please go ahead and feel free to apply the modified patch.

Thanks a lot for your help.
Regards,
Bhupesh

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#4 lore raw

RE: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Hi Bhupesh,

Quoted 153 lines · Bhupesh Sharma

Hello Kazu,

Many thanks for your review comments.

Quoted 148 lines · Kazuhito Hagio
Quoted 71 lines · Bhupesh Sharma

The existing methodology to obtain 'info->page_offset' from reading _stext symbol (from kallsyms) doesn't work well in KASLR boot cases on arm64 machines as the PAGE_OFFSET (or the virtual address which indicates the start of the linear region) can be randomized as well on basis of the kaslr-seed.

Since the value of PAGE_OFFSET inside the kernel is randomized in such cases and there is no existing mechanism of conveying this value from kernel-space to user-space, so we can use the method used by archs like x86_64 to generate the 'info->page_offset' value from the PT_LOAD segments by subtracting the phy_addr from virt_addr of a PT_LOAD segment.

This approach works fine both with KASLR and non-KASLR boot cases.

I tested this on my qualcomm-amberwing board. Here are some logs from the KASLR boot cases:

- Verify that the EFI firmware supports 'kaslr-seed':

      chosen {
              kaslr-seed = <0x0 0x0>;
              <..snip..>
      };

- Verify that '--mem-usage' works well after this fix as well (I used kernel 4.18.0-rc4+ for my checks):

The kernel version is not supported.
The makedumpfile operation may be incomplete.

TYPE PAGES EXCLUDABLE DESCRIPTION
----------------------------------------------------------------------
ZERO 4396 yes Pages filled
with zero
NON_PRI_CACHE 27859 yes Cache pages
without private flag
PRI_CACHE 18490 yes Cache pages with
private flag
USER 2728 yes User process
pages

FREE          1465848                 yes             Free pages
KERN_DATA     18537                   no              Dumpable kernel

data

page size:            65536
Total pages on system:        1537858
Total size on system: 100785061888     Byte
---
 arch/arm64.c   | 23 ++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  1 +
 3 files changed, 20 insertions(+), 5 deletions(-)
arch/arm64.c +3 -0
index 2fd3e1874376..9e8c77c76935 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+     int i;
+     unsigned long long phys_start;
+     unsigned long long virt_start;
      ulong _stext;
 
      _stext = get_stext_symbol();

arch/arm64.c get_versiondep_info_arm64(void) · line 296

Quoted 8 lines
arch/arm64.c +0 -4
@@ -289,12 +292,22 @@ get_versiondep_info_arm64(void)
              return FALSE;
      }
 
 
-     info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
-
-     DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
-                     va_bits);

According to makedumpfile commit 4944f93484 ("x86_64: Calculate page_offset in case of re-filtering/sadump/virsh dump"), in case of re-filtering, we don't have any PT_LOAD. So, with this patch, we cannot re-filter dump file on arm64 system?

If we cannot, is it better to leave the above behind the following pt_load section for re-filtering non-KASLR dump?

Quoted 4 lines
arch/arm64.c +4 -0
+     if (get_num_pt_loads()) {
+             for (i = 0;
+                     get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+                     i++) {

arch/arm64.c get_versiondep_info_arm64(void) · line 309

Quoted 10 lines
arch/arm64.c +10 -0
+                     if (virt_start != NOT_KV_ADDR
+                                     && virt_start < __START_KERNEL_map
+                                     && phys_start != NOT_PADDR && phys_start != NOT_PADDR_ARM64) {
+                             info->page_offset = virt_start - phys_start;
+                             DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+                                             info->page_offset, va_bits);
+                             return TRUE;
+                     }
+             }
+     }

I'll adjust some indents for readability.

Quoted 11 lines
arch/arm64.c +1 -1
-     return TRUE;
+     return FALSE;
 }
 
 /*
common.h +0 -0
index 6e2f657a79c7..a8181777dbb7 100644

common.h line 52

Quoted 5 lines
common.h +1 -0
@@ -48,6 +48,7 @@
 #define NOT_MEMMAP_ADDR      (0x0)
 #define NOT_KV_ADDR  (0x0)
 #define NOT_PADDR    (ULONGLONG_MAX)
 
+#define NOT_PADDR_ARM64      (0x0000000010a80000UL)

I think that this should not be in common.h, because it's not for arch-specific definitions, so I'll move it to makedumpfile.h.

makedumpfile.h struct domain_list { · line 2023

Quoted 12 lines
common.h +0 -0
 #define BADADDR      ((ulong)(-1))
 

#endif /* COMMON_H */

makedumpfile.h +1 -0
index 5ff94b8e4ac6..5297279f0f3b 100644
@@ -2020,6 +2020,7 @@ struct domain_list {
 #define MFNS_PER_FRAME               (info->page_size / sizeof(unsigned long))
 
 #ifdef __aarch64__
+#define __START_KERNEL_map           (0xffffffff80000000UL)

This #ifdef here is for the definitions related to Xen extraction. I'll move it to the same place as the above.

Quoted 4 lines
makedumpfile.h +0 -0
 unsigned long long kvtop_xen_arm64(unsigned long kvaddr);
 #define kvtop_xen(X) kvtop_xen_arm64(X)
 #endif /* aarch64 */
Quoted 5 lines
Quoted 2 lines

I attached a patch that was applied my suggestions. How about it?

I agree. The patch looks good with the suggested changes. Please go ahead and feel free to apply the modified patch.

OK, I'll apply it.

With respect to the 2/2 patch, I'm waiting for the kernel patch to be merged.

If need be, I can apply the 1/2 patch separately from the 2/2 patch, because I think they are not directly connected with each other.

Thanks,
Kazu

p.s. it seems that I cannot send emails to gmail for now, so dropped your gmail address. sorry for the inconvenience.

Quoted 5 lines

Thanks a lot for your help.
Regards,
Bhupesh

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#5 lore raw

Re: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Hi Kazu,

Quoted 162 lines · Kazuhito Hagio

Hi Bhupesh,

Quoted 153 lines · Bhupesh Sharma

Hello Kazu,

Many thanks for your review comments.

Quoted 148 lines · Kazuhito Hagio
Quoted 71 lines · Bhupesh Sharma

The existing methodology to obtain 'info->page_offset' from reading _stext symbol (from kallsyms) doesn't work well in KASLR boot cases on arm64 machines as the PAGE_OFFSET (or the virtual address which indicates the start of the linear region) can be randomized as well on basis of the kaslr-seed.

Since the value of PAGE_OFFSET inside the kernel is randomized in such cases and there is no existing mechanism of conveying this value from kernel-space to user-space, so we can use the method used by archs like x86_64 to generate the 'info->page_offset' value from the PT_LOAD segments by subtracting the phy_addr from virt_addr of a PT_LOAD segment.

This approach works fine both with KASLR and non-KASLR boot cases.

I tested this on my qualcomm-amberwing board. Here are some logs from the KASLR boot cases:

- Verify that the EFI firmware supports 'kaslr-seed':

      chosen {
              kaslr-seed = <0x0 0x0>;
              <..snip..>
      };

- Verify that '--mem-usage' works well after this fix as well (I used kernel 4.18.0-rc4+ for my checks):

The kernel version is not supported.
The makedumpfile operation may be incomplete.

TYPE PAGES EXCLUDABLE DESCRIPTION
----------------------------------------------------------------------
ZERO 4396 yes Pages filled
with zero
NON_PRI_CACHE 27859 yes Cache pages
without private flag
PRI_CACHE 18490 yes Cache pages with
private flag
USER 2728 yes User process
pages

FREE          1465848                 yes             Free pages
KERN_DATA     18537                   no              Dumpable kernel

data

page size:            65536
Total pages on system:        1537858
Total size on system: 100785061888     Byte
---
 arch/arm64.c   | 23 ++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  1 +
 3 files changed, 20 insertions(+), 5 deletions(-)
arch/arm64.c +3 -0
index 2fd3e1874376..9e8c77c76935 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+     int i;
+     unsigned long long phys_start;
+     unsigned long long virt_start;
      ulong _stext;
 
      _stext = get_stext_symbol();

arch/arm64.c get_versiondep_info_arm64(void) · line 296

Quoted 8 lines
arch/arm64.c +0 -4
@@ -289,12 +292,22 @@ get_versiondep_info_arm64(void)
              return FALSE;
      }
 
 
-     info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
-
-     DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
-                     va_bits);

According to makedumpfile commit 4944f93484 ("x86_64: Calculate page_offset in case of re-filtering/sadump/virsh dump"), in case of re-filtering, we don't have any PT_LOAD. So, with this patch, we cannot re-filter dump file on arm64 system?

If we cannot, is it better to leave the above behind the following pt_load section for re-filtering non-KASLR dump?

Quoted 4 lines
arch/arm64.c +4 -0
+     if (get_num_pt_loads()) {
+             for (i = 0;
+                     get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+                     i++) {

arch/arm64.c get_versiondep_info_arm64(void) · line 309

Quoted 10 lines
arch/arm64.c +10 -0
+                     if (virt_start != NOT_KV_ADDR
+                                     && virt_start < __START_KERNEL_map
+                                     && phys_start != NOT_PADDR && phys_start != NOT_PADDR_ARM64) {
+                             info->page_offset = virt_start - phys_start;
+                             DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+                                             info->page_offset, va_bits);
+                             return TRUE;
+                     }
+             }
+     }

I'll adjust some indents for readability.

Quoted 11 lines
arch/arm64.c +1 -1
-     return TRUE;
+     return FALSE;
 }
 
 /*
common.h +0 -0
index 6e2f657a79c7..a8181777dbb7 100644

common.h line 52

Quoted 5 lines
common.h +1 -0
@@ -48,6 +48,7 @@
 #define NOT_MEMMAP_ADDR      (0x0)
 #define NOT_KV_ADDR  (0x0)
 #define NOT_PADDR    (ULONGLONG_MAX)
 
+#define NOT_PADDR_ARM64      (0x0000000010a80000UL)

I think that this should not be in common.h, because it's not for arch-specific definitions, so I'll move it to makedumpfile.h.

makedumpfile.h struct domain_list { · line 2023

Quoted 12 lines
common.h +0 -0
 #define BADADDR      ((ulong)(-1))
 

#endif /* COMMON_H */

makedumpfile.h +1 -0
index 5ff94b8e4ac6..5297279f0f3b 100644
@@ -2020,6 +2020,7 @@ struct domain_list {
 #define MFNS_PER_FRAME               (info->page_size / sizeof(unsigned long))
 
 #ifdef __aarch64__
+#define __START_KERNEL_map           (0xffffffff80000000UL)

This #ifdef here is for the definitions related to Xen extraction. I'll move it to the same place as the above.

Quoted 4 lines
makedumpfile.h +0 -0
 unsigned long long kvtop_xen_arm64(unsigned long kvaddr);
 #define kvtop_xen(X) kvtop_xen_arm64(X)
 #endif /* aarch64 */
Quoted 5 lines
Quoted 2 lines

I attached a patch that was applied my suggestions. How about it?

I agree. The patch looks good with the suggested changes. Please go ahead and feel free to apply the modified patch.

Quoted 7 lines

OK, I'll apply it.

With respect to the 2/2 patch, I'm waiting for the kernel patch to be merged.

If need be, I can apply the 1/2 patch separately from the 2/2 patch, because I think they are not directly connected with each other.

Sure, I think we can handle 2/2 patch when it's kernel part is reviewed properly by the arm64 kernel maintainer(s). We can apply 1/2 patch in the meanwhile.

Quoted 2 lines

p.s. it seems that I cannot send emails to gmail for now, so dropped your gmail address. sorry for the inconvenience.

Sure, I use the gmail address just as a backup to read/answer upstream replies once I am out of office (as it is easier to access on my phone).

Thanks,
Bhupesh

Quoted 5 lines
Quoted 5 lines

Thanks a lot for your help.
Regards,
Bhupesh

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#6 lore raw

Re: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Hi Kazu,

Quoted 180 lines · Bhupesh Sharma

Hi Kazu,

Quoted 162 lines · Kazuhito Hagio

Hi Bhupesh,

Quoted 153 lines · Bhupesh Sharma

Hello Kazu,

Many thanks for your review comments.

Quoted 148 lines · Kazuhito Hagio
Quoted 71 lines · Bhupesh Sharma

The existing methodology to obtain 'info->page_offset' from reading _stext symbol (from kallsyms) doesn't work well in KASLR boot cases on arm64 machines as the PAGE_OFFSET (or the virtual address which indicates the start of the linear region) can be randomized as well on basis of the kaslr-seed.

Since the value of PAGE_OFFSET inside the kernel is randomized in such cases and there is no existing mechanism of conveying this value from kernel-space to user-space, so we can use the method used by archs like x86_64 to generate the 'info->page_offset' value from the PT_LOAD segments by subtracting the phy_addr from virt_addr of a PT_LOAD segment.

This approach works fine both with KASLR and non-KASLR boot cases.

I tested this on my qualcomm-amberwing board. Here are some logs from the KASLR boot cases:

- Verify that the EFI firmware supports 'kaslr-seed':

      chosen {
              kaslr-seed = <0x0 0x0>;
              <..snip..>
      };

- Verify that '--mem-usage' works well after this fix as well (I used kernel 4.18.0-rc4+ for my checks):

The kernel version is not supported.
The makedumpfile operation may be incomplete.

TYPE PAGES EXCLUDABLE DESCRIPTION
----------------------------------------------------------------------
ZERO 4396 yes Pages filled
with zero
NON_PRI_CACHE 27859 yes Cache pages
without private flag
PRI_CACHE 18490 yes Cache pages with
private flag
USER 2728 yes User process
pages

FREE          1465848                 yes             Free pages
KERN_DATA     18537                   no              Dumpable kernel

data

page size:            65536
Total pages on system:        1537858
Total size on system: 100785061888     Byte
---
 arch/arm64.c   | 23 ++++++++++++++++++-----
 common.h       |  1 +
 makedumpfile.h |  1 +
 3 files changed, 20 insertions(+), 5 deletions(-)
arch/arm64.c +3 -0
index 2fd3e1874376..9e8c77c76935 100644
@@ -265,6 +265,9 @@ get_xen_info_arm64(void)
 int
 get_versiondep_info_arm64(void)
 {
+     int i;
+     unsigned long long phys_start;
+     unsigned long long virt_start;
      ulong _stext;
 
      _stext = get_stext_symbol();

arch/arm64.c get_versiondep_info_arm64(void) · line 296

Quoted 8 lines
arch/arm64.c +0 -4
@@ -289,12 +292,22 @@ get_versiondep_info_arm64(void)
              return FALSE;
      }
 
 
-     info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
-
-     DEBUG_MSG("page_offset=%lx, va_bits=%d\n", info->page_offset,
-                     va_bits);

According to makedumpfile commit 4944f93484 ("x86_64: Calculate page_offset in case of re-filtering/sadump/virsh dump"), in case of re-filtering, we don't have any PT_LOAD. So, with this patch, we cannot re-filter dump file on arm64 system?

If we cannot, is it better to leave the above behind the following pt_load section for re-filtering non-KASLR dump?

Quoted 4 lines
arch/arm64.c +4 -0
+     if (get_num_pt_loads()) {
+             for (i = 0;
+                     get_pt_load(i, &phys_start, NULL, &virt_start, NULL);
+                     i++) {

arch/arm64.c get_versiondep_info_arm64(void) · line 309

Quoted 10 lines
arch/arm64.c +10 -0
+                     if (virt_start != NOT_KV_ADDR
+                                     && virt_start < __START_KERNEL_map
+                                     && phys_start != NOT_PADDR && phys_start != NOT_PADDR_ARM64) {
+                             info->page_offset = virt_start - phys_start;
+                             DEBUG_MSG("info->page_offset: %lx, VA_BITS: %d\n",
+                                             info->page_offset, va_bits);
+                             return TRUE;
+                     }
+             }
+     }

I'll adjust some indents for readability.

Quoted 11 lines
arch/arm64.c +1 -1
-     return TRUE;
+     return FALSE;
 }
 
 /*
common.h +0 -0
index 6e2f657a79c7..a8181777dbb7 100644

common.h line 52

Quoted 5 lines
common.h +1 -0
@@ -48,6 +48,7 @@
 #define NOT_MEMMAP_ADDR      (0x0)
 #define NOT_KV_ADDR  (0x0)
 #define NOT_PADDR    (ULONGLONG_MAX)
 
+#define NOT_PADDR_ARM64      (0x0000000010a80000UL)

I think that this should not be in common.h, because it's not for arch-specific definitions, so I'll move it to makedumpfile.h.

makedumpfile.h struct domain_list { · line 2023

Quoted 12 lines
common.h +0 -0
 #define BADADDR      ((ulong)(-1))
 

#endif /* COMMON_H */

makedumpfile.h +1 -0
index 5ff94b8e4ac6..5297279f0f3b 100644
@@ -2020,6 +2020,7 @@ struct domain_list {
 #define MFNS_PER_FRAME               (info->page_size / sizeof(unsigned long))
 
 #ifdef __aarch64__
+#define __START_KERNEL_map           (0xffffffff80000000UL)

This #ifdef here is for the definitions related to Xen extraction. I'll move it to the same place as the above.

Quoted 4 lines
makedumpfile.h +0 -0
 unsigned long long kvtop_xen_arm64(unsigned long kvaddr);
 #define kvtop_xen(X) kvtop_xen_arm64(X)
 #endif /* aarch64 */
Quoted 5 lines
Quoted 2 lines

I attached a patch that was applied my suggestions. How about it?

I agree. The patch looks good with the suggested changes. Please go ahead and feel free to apply the modified patch.

Quoted 7 lines

OK, I'll apply it.

With respect to the 2/2 patch, I'm waiting for the kernel patch to be merged.

If need be, I can apply the 1/2 patch separately from the 2/2 patch, because I think they are not directly connected with each other.

Sure, I think we can handle 2/2 patch when it's kernel part is reviewed properly by the arm64 kernel maintainer(s). We can apply 1/2 patch in the meanwhile.

Quoted 2 lines

p.s. it seems that I cannot send emails to gmail for now, so dropped your gmail address. sorry for the inconvenience.

Quoted 6 lines

Sure, I use the gmail address just as a backup to read/answer upstream replies once I am out of office (as it is easier to access on my phone).

Thanks,
Bhupesh

The kernel patch is now accepted in Linus's tree:
commit e401b7c2c69008ad2fcdc154f7c5421281c90042

Author: Bhupesh Sharma <[email protected]>
Date:   Mon Jul 30 11:54:43 2018 +0530

arm64, kaslr: export offset in VMCOREINFO ELF notes

Can you please pick this patchset and now apply it to makedumpfile source?

Thanks,
Bhupesh

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#7 lore raw

RE: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Hello Bhupesh,

Quoted 8 lines · Bhupesh Sharma

The kernel patch is now accepted in Linus's tree:
commit e401b7c2c69008ad2fcdc154f7c5421281c90042

Author: Bhupesh Sharma <[email protected]>
Date:   Mon Jul 30 11:54:43 2018 +0530

arm64, kaslr: export offset in VMCOREINFO ELF notes

Can you please pick this patchset and now apply it to makedumpfile source?

Thank you for informing me about it, and sorry for my delayed response. I applied the patches to the devel branch.

commit 616c98d1abac6f348f13cfe55e2df5e1d5d26781

Author: Bhupesh Sharma <[email protected]>
Date:   Thu Jul 19 11:13:41 2018 +0530

[PATCH] arm64: Add runtime kaslr offset if it exists

commit 94c97db3fe859ca14d7b38b0ae9ee0ffb83689d2

Author: Bhupesh Sharma <[email protected]>
Date:   Thu Jul 19 11:13:40 2018 +0530
    [PATCH] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Thanks,
Kazu

Quoted 3 lines

Thanks,
Bhupesh

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#8 lore raw

Re: [PATCH 1/2] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Hi Kazu,

Quoted 19 lines · Kazuhito Hagio

Hello Bhupesh,

Quoted 8 lines · Bhupesh Sharma

The kernel patch is now accepted in Linus's tree:
commit e401b7c2c69008ad2fcdc154f7c5421281c90042

Author: Bhupesh Sharma <[email protected]>
Date:   Mon Jul 30 11:54:43 2018 +0530

arm64, kaslr: export offset in VMCOREINFO ELF notes

Can you please pick this patchset and now apply it to makedumpfile source?

Thank you for informing me about it, and sorry for my delayed response. I applied the patches to the devel branch.

commit 616c98d1abac6f348f13cfe55e2df5e1d5d26781

Author: Bhupesh Sharma <[email protected]>
Date:   Thu Jul 19 11:13:41 2018 +0530
Quoted 8 lines

[PATCH] arm64: Add runtime kaslr offset if it exists

commit 94c97db3fe859ca14d7b38b0ae9ee0ffb83689d2

Author: Bhupesh Sharma <[email protected]>
Date:   Thu Jul 19 11:13:40 2018 +0530
    [PATCH] arm64: Get 'info->page_offset' from PT_LOAD segments to support KASLR boot cases

Thanks. I have sent out another fix for 94c97db3fe859ca14d7b38b0ae9ee0ffb83689d2 here
<http://lists.infradead.org/pipermail/kexec/2018-August/021390.html>,
which removes the
hardcoded NOT_PADDR_ARM64 macro value and makes it platform independent.

Kindly review the same and consider it applying to the makedumpfile code as well.

Regards,
Bhupesh

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#9 lore raw

PATCH 2/2 arm64: Add runtime kaslr offset if it exists

In reply to Bhupesh Sharma

If we have to erase a symbol from vmcore whose address is not present in vmcoreinfo, then we need to pass vmlinux as well to get the symbol address.
When kaslr is enabled, virtual address of all the kernel symbols are randomized with an offset. vmlinux always has a static address, but all the arch specific calculation are based on run time kernel address. So we need to find a way to translate symbol address from vmlinux to kernel run time address.

Without this patch:

   # cat > scrub.conf << EOF
   [vmlinux]
   erase jiffies
   erase init_task.utime
   for tsk in init_task.tasks.next within task_struct:tasks
       erase tsk.utime
   endfor
   EOF
  # makedumpfile --split -d 31 -x vmlinux --config scrub.conf vmcore dumpfile_{1,2,3}
  readpage_elf: Attempt to read non-existent page at 0xffffa8a5bf180000.
  readmem: type_addr: 1, addr:ffffa8a5bf180000, size:8
  vaddr_to_paddr_arm64: Can't read pgd
  readmem: Can't convert a virtual address(ffff0000092a542c) to physical
  address.
  readmem: type_addr: 0, addr:ffff0000092a542c, size:390
  check_release: Can't get the address of system_utsname

After this patch check_release() is ok, and also we are able to erase symbol from vmcore (I checked this with kernel 4.18.0-rc4+):

  # makedumpfile --split -d 31 -x vmlinux --config scrub.conf vmcore dumpfile_{1,2,3}
  The kernel version is not supported.
  The makedumpfile operation may be incomplete.
  Checking for memory holes                         : [100.0 %] \
  Checking for memory holes                         : [100.0 %] |
  Checking foExcluding unnecessary pages                       : [100.0 %]
  \
  Excluding unnecessary pages                       : [100.0 %] \

The dumpfiles are saved to dumpfile_1, dumpfile_2, and dumpfile_3.

makedumpfile Completed.

This feature also requires a fix in the kernel as well which has been submitted upstream (see[0]).

[0] https://patchwork.kernel.org/patch/10533333/

---
 arch/arm64.c   | 41 +++++++++++++++++++++++++++++++++++++++++
 makedumpfile.h |  3 ++-
 2 files changed, 43 insertions(+), 1 deletion(-)
arch/arm64.c +41 -0
index 9e8c77c76935..9772022605ec 100644
@@ -181,6 +181,47 @@ get_phys_base_arm64(void)
 	return TRUE;
 }
 
+unsigned long
+get_kaslr_offset_arm64(unsigned long vaddr)
+{
+	unsigned int i;
+	char buf[BUFSIZE_FGETS], *endp;
+
+	if (!info->kaslr_offset && info->file_vmcoreinfo) {
+		if (fseek(info->file_vmcoreinfo, 0, SEEK_SET) < 0) {
+			ERRMSG("Can't seek the vmcoreinfo file(%s). %s\n",
+					info->name_vmcoreinfo, strerror(errno));
+			return FALSE;
+		}
+
+		while (fgets(buf, BUFSIZE_FGETS, info->file_vmcoreinfo)) {
+			i = strlen(buf);
+			if (!i)
+				break;
+			if (buf[i - 1] == '\n')
+				buf[i - 1] = '\0';
+			if (strncmp(buf, STR_KERNELOFFSET,
+					strlen(STR_KERNELOFFSET)) == 0) {
+				info->kaslr_offset =
+					strtoul(buf+strlen(STR_KERNELOFFSET),&endp,16);
+				DEBUG_MSG("info->kaslr_offset: %lx\n", info->kaslr_offset);
+			}
+		}
+	}
+
+	if (vaddr >= __START_KERNEL_map &&
+			vaddr < __START_KERNEL_map + info->kaslr_offset) {
+		DEBUG_MSG("info->kaslr_offset: %lx\n", info->kaslr_offset);
+		return info->kaslr_offset;
+	} else {
+		/*
+		 * TODO: we need to check if it is vmalloc/vmmemmap/module
+		 * address, we will have different offset
+		 */
+		return 0;
+	}
+}
+
 ulong
 get_stext_symbol(void)
 {
makedumpfile.h +2 -1
index 5297279f0f3b..4daf71ef14b7 100644
@@ -967,12 +967,13 @@ unsigned long long vaddr_to_paddr_arm64(unsigned long vaddr);
 int get_versiondep_info_arm64(void);
 int get_xen_basic_info_arm64(void);
 int get_xen_info_arm64(void);
+unsigned long get_kaslr_offset_arm64(unsigned long vaddr);
 #define find_vmemmap()		stub_false()
 #define vaddr_to_paddr(X)	vaddr_to_paddr_arm64(X)
 #define get_phys_base()		get_phys_base_arm64()
 #define get_machdep_info()	get_machdep_info_arm64()
 #define get_versiondep_info()	get_versiondep_info_arm64()
-#define get_kaslr_offset(X)	stub_false()
+#define get_kaslr_offset(X)	get_kaslr_offset_arm64(X)
 #define get_xen_basic_info_arch(X) get_xen_basic_info_arm64(X)
 #define get_xen_info_arch(X) get_xen_info_arm64(X)
 #define is_phys_addr(X)		stub_true_ul(X)
Signature · 7 lines
2.7.4


_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec