Skip to contentSkip to conversation

efi/libstub/arm64: Report meaningful relocation errors

PATCH In review Kees Cook 3 participants 7 messages Aug 14, 2019 – Sep 25, 2019 EFI

lore
Share & tools
lore

Kees Cook

author
#0 lore raw

PATCH 1/1 efi/libstub/arm64: Report meaningful relocation errors

When UEFI booting, if allocate_pages() fails (either via KASLR or regular boot), efi_low_alloc() is used for fall back. If it, too, fails, it reports "Failed to relocate kernel". Then handle_kernel_image() reports the failure to its caller, which unhelpfully reports exactly the same string again:

EFI stub: ERROR: Failed to relocate kernel
EFI stub: ERROR: Failed to relocate kernel

While debugging linker errors in the UEFI code that created insane memory sizes that all the allocation attempts would fail at, this was a cause for confusion. Knowing each allocation had failed would have helped me isolate the issue sooner. To that end, this improves the error messages to detail which specific allocations have failed.

---
 drivers/firmware/efi/libstub/arm64-stub.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/firmware/efi/libstub/arm64-stub.c +5 -1
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 		status = efi_random_alloc(sys_table_arg, *reserve_size,
 					  MIN_KIMG_ALIGN, reserve_addr,
 					  (u32)phys_seed);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
 		*image_addr = *reserve_addr + offset;
 	} else {
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 					EFI_LOADER_DATA,
 					*reserve_size / EFI_PAGE_SIZE,
 					(efi_physical_addr_t *)reserve_addr);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
 	}
 
 	if (status != EFI_SUCCESS) {
@@ -143,7 +147,7 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 				       MIN_KIMG_ALIGN, reserve_addr);
 
 		if (status != EFI_SUCCESS) {
-			pr_efi_err(sys_table_arg, "Failed to relocate kernel\n");
+			pr_efi_err(sys_table_arg, "efi_low_alloc() failed\n");
 			*reserve_size = 0;
 			return status;
 		}
Signature · 5 lines
2.17.1


-- 
Kees Cook
#1 lore raw

Hi Kees,

Quoted 34 lines · Kees Cook

When UEFI booting, if allocate_pages() fails (either via KASLR or regular boot), efi_low_alloc() is used for fall back. If it, too, fails, it reports "Failed to relocate kernel". Then handle_kernel_image() reports the failure to its caller, which unhelpfully reports exactly the same string again:

EFI stub: ERROR: Failed to relocate kernel
EFI stub: ERROR: Failed to relocate kernel

While debugging linker errors in the UEFI code that created insane memory sizes that all the allocation attempts would fail at, this was a cause for confusion. Knowing each allocation had failed would have helped me isolate the issue sooner. To that end, this improves the error messages to detail which specific allocations have failed.

---
 drivers/firmware/efi/libstub/arm64-stub.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 		status = efi_random_alloc(sys_table_arg, *reserve_size,
 					  MIN_KIMG_ALIGN, reserve_addr,
 					  (u32)phys_seed);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
 		*image_addr = *reserve_addr + offset;
 	} else {

drivers/firmware/efi/libstub/arm64-stub.c efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg, · line 142

Quoted 6 lines
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 					EFI_LOADER_DATA,
 					*reserve_size / EFI_PAGE_SIZE,
 					(efi_physical_addr_t *)reserve_addr);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
 	}

Not sure I see the need to distinsuish the 'KASLR' case from the 'regular' case -- only one should run, right? That also didn't seem to be part of the use-case in the commit, unless I'm missing something.

Maybe combine the prints as per the diff below?

Will

--->8

drivers/firmware/efi/libstub/arm64-stub.c +4 -2
index 1550d244e996..820c58cc149e 100644
@@ -143,13 +143,15 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 				       MIN_KIMG_ALIGN, reserve_addr);
 
 		if (status != EFI_SUCCESS) {
-			pr_efi_err(sys_table_arg, "Failed to relocate kernel\n");
+			pr_efi_err(sys_table_arg, "efi_low_alloc() failed\n");
 			*reserve_size = 0;
 			return status;
 		}
 		*image_addr = *reserve_addr + TEXT_OFFSET;
+	} else {
+		pr_efi_err(sys_table_arg, "allocate_pages() failed\n");
 	}
-	memcpy((void *)*image_addr, old_image_addr, kernel_size);
 
+	memcpy((void *)*image_addr, old_image_addr, kernel_size);
 	return EFI_SUCCESS;
 }
author
#2 lore raw
Quoted 34 lines · Will Deacon

Hi Kees,

Quoted 31 lines · Kees Cook

When UEFI booting, if allocate_pages() fails (either via KASLR or regular boot), efi_low_alloc() is used for fall back. If it, too, fails, it reports "Failed to relocate kernel". Then handle_kernel_image() reports the failure to its caller, which unhelpfully reports exactly the same string again:

EFI stub: ERROR: Failed to relocate kernel
EFI stub: ERROR: Failed to relocate kernel

While debugging linker errors in the UEFI code that created insane memory sizes that all the allocation attempts would fail at, this was a cause for confusion. Knowing each allocation had failed would have helped me isolate the issue sooner. To that end, this improves the error messages to detail which specific allocations have failed.

---
 drivers/firmware/efi/libstub/arm64-stub.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 		status = efi_random_alloc(sys_table_arg, *reserve_size,
 					  MIN_KIMG_ALIGN, reserve_addr,
 					  (u32)phys_seed);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
Quoted 13 lines

drivers/firmware/efi/libstub/arm64-stub.c efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg, · line 142

Quoted 9 lines
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
 		*image_addr = *reserve_addr + offset;
 	} else {
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 					EFI_LOADER_DATA,
 					*reserve_size / EFI_PAGE_SIZE,
 					(efi_physical_addr_t *)reserve_addr);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
 	}

Not sure I see the need to distinsuish the 'KASLR' case from the 'regular' case -- only one should run, right? That also didn't seem to be part of the use-case in the commit, unless I'm missing something.

I just did that to help with differentiating the cases. Maybe something was special about KASLR picking the wrong location that triggered the failure, etc.

Quoted 1 line

Maybe combine the prints as per the diff below?

That could work. If you're against the KASLR vs regular thing, I can respin the patch?

-Kees

Quoted 27 lines

Will

--->8

drivers/firmware/efi/libstub/arm64-stub.c +4 -2
index 1550d244e996..820c58cc149e 100644
@@ -143,13 +143,15 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 				       MIN_KIMG_ALIGN, reserve_addr);
 
 		if (status != EFI_SUCCESS) {
-			pr_efi_err(sys_table_arg, "Failed to relocate kernel\n");
+			pr_efi_err(sys_table_arg, "efi_low_alloc() failed\n");
 			*reserve_size = 0;
 			return status;
 		}
 		*image_addr = *reserve_addr + TEXT_OFFSET;
+	} else {
+		pr_efi_err(sys_table_arg, "allocate_pages() failed\n");
 	}
-	memcpy((void *)*image_addr, old_image_addr, kernel_size);
 
+	memcpy((void *)*image_addr, old_image_addr, kernel_size);
 	return EFI_SUCCESS;
 }
Signature · 1 line
Kees Cook
#3 lore raw
Quoted 27 lines · Kees Cook
Quoted 12 lines · Will Deacon
Quoted 11 lines · Kees Cook
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 		status = efi_random_alloc(sys_table_arg, *reserve_size,
 					  MIN_KIMG_ALIGN, reserve_addr,
 					  (u32)phys_seed);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
Quoted 13 lines

drivers/firmware/efi/libstub/arm64-stub.c efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg, · line 142

Quoted 9 lines
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
 		*image_addr = *reserve_addr + offset;
 	} else {
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 					EFI_LOADER_DATA,
 					*reserve_size / EFI_PAGE_SIZE,
 					(efi_physical_addr_t *)reserve_addr);
+		if (status != EFI_SUCCESS)
+			pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
 	}

Not sure I see the need to distinsuish the 'KASLR' case from the 'regular' case -- only one should run, right? That also didn't seem to be part of the use-case in the commit, unless I'm missing something.

Quoted 8 lines

I just did that to help with differentiating the cases. Maybe something was special about KASLR picking the wrong location that triggered the failure, etc.

Quoted 1 line

Maybe combine the prints as per the diff below?

That could work. If you're against the KASLR vs regular thing, I can respin the patch?

Happy to Ack it with that change, although I suppose it's ultimately up to Ard :)

Will

#4 lore raw
Quoted 33 lines · Will Deacon
Quoted 31 lines · Kees Cook
Quoted 12 lines · Will Deacon
Quoted 11 lines · Kees Cook
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
          status = efi_random_alloc(sys_table_arg, *reserve_size,
                                    MIN_KIMG_ALIGN, reserve_addr,
                                    (u32)phys_seed);
+         if (status != EFI_SUCCESS)
+                 pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
Quoted 13 lines

drivers/firmware/efi/libstub/arm64-stub.c efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg, · line 142

Quoted 9 lines
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
          *image_addr = *reserve_addr + offset;
  } else {
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
                                  EFI_LOADER_DATA,
                                  *reserve_size / EFI_PAGE_SIZE,
                                  (efi_physical_addr_t *)reserve_addr);
+         if (status != EFI_SUCCESS)
+                 pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
  }

Not sure I see the need to distinsuish the 'KASLR' case from the 'regular' case -- only one should run, right? That also didn't seem to be part of the use-case in the commit, unless I'm missing something.

I just did that to help with differentiating the cases. Maybe something was special about KASLR picking the wrong location that triggered the failure, etc.

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

Maybe combine the prints as per the diff below?

That could work. If you're against the KASLR vs regular thing, I can respin the patch?

Happy to Ack it with that change, although I suppose it's ultimately up to Ard :)

No objections from me, but I prefer Will's version.

author
#5 lore raw
Quoted 36 lines · Ard Biesheuvel
Quoted 35 lines · Will Deacon
Quoted 33 lines · Kees Cook
Quoted 12 lines · Will Deacon
Quoted 11 lines · Kees Cook
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
          status = efi_random_alloc(sys_table_arg, *reserve_size,
                                    MIN_KIMG_ALIGN, reserve_addr,
                                    (u32)phys_seed);
+         if (status != EFI_SUCCESS)
+                 pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
Quoted 13 lines

drivers/firmware/efi/libstub/arm64-stub.c efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg, · line 142

Quoted 9 lines
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
          *image_addr = *reserve_addr + offset;
  } else {
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
                                  EFI_LOADER_DATA,
                                  *reserve_size / EFI_PAGE_SIZE,
                                  (efi_physical_addr_t *)reserve_addr);
+         if (status != EFI_SUCCESS)
+                 pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
  }

Not sure I see the need to distinsuish the 'KASLR' case from the 'regular' case -- only one should run, right? That also didn't seem to be part of the use-case in the commit, unless I'm missing something.

I just did that to help with differentiating the cases. Maybe something was special about KASLR picking the wrong location that triggered the failure, etc.

Quoted 1 line

Maybe combine the prints as per the diff below?

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

That could work. If you're against the KASLR vs regular thing, I can respin the patch?

Happy to Ack it with that change, although I suppose it's ultimately up to Ard :)

No objections from me, but I prefer Will's version.

I took a look at this again... to report the failures as Will suggests, it would look like this:

drivers/firmware/efi/libstub/arm64-stub.c +3 -1
@@ -138,12 +138,14 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
 	}
 
 	if (status != EFI_SUCCESS) {
+		pr_efi_err(sys_table_arg, "allocate_pages() failed\n");
+
 		*reserve_size = kernel_memsize + TEXT_OFFSET;
 		status = efi_low_alloc(sys_table_arg, *reserve_size,
 				       MIN_KIMG_ALIGN, reserve_addr);
 
 		if (status != EFI_SUCCESS) {
-			pr_efi_err(sys_table_arg, "Failed to relocate kernel\n");
+			pr_efi_err(sys_table_arg, "efi_low_alloc() failed\n");
 			*reserve_size = 0;
 			return status;
 		}

My reasoning for putting the failure earlier is to differentiate which path was taken where the allocate_pages() failed: either regular or KASLR. If that's really not considered important here, I can send the above patch... Thoughts?

Signature · 1 line
Kees Cook
#6 lore raw
Quoted 69 lines · Kees Cook
Quoted 36 lines · Ard Biesheuvel
Quoted 35 lines · Will Deacon
Quoted 33 lines · Kees Cook
Quoted 12 lines · Will Deacon
Quoted 11 lines · Kees Cook
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
index 1550d244e996..24022f956e01 100644
@@ -111,6 +111,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
          status = efi_random_alloc(sys_table_arg, *reserve_size,
                                    MIN_KIMG_ALIGN, reserve_addr,
                                    (u32)phys_seed);
+         if (status != EFI_SUCCESS)
+                 pr_efi_err(sys_table_arg, "KASLR allocate_pages() failed\n");
 
Quoted 13 lines

drivers/firmware/efi/libstub/arm64-stub.c efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg, · line 142

Quoted 9 lines
drivers/firmware/efi/libstub/arm64-stub.c +2 -0
          *image_addr = *reserve_addr + offset;
  } else {
@@ -135,6 +137,8 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
                                  EFI_LOADER_DATA,
                                  *reserve_size / EFI_PAGE_SIZE,
                                  (efi_physical_addr_t *)reserve_addr);
+         if (status != EFI_SUCCESS)
+                 pr_efi_err(sys_table_arg, "regular allocate_pages() failed\n");
  }

Not sure I see the need to distinsuish the 'KASLR' case from the 'regular' case -- only one should run, right? That also didn't seem to be part of the use-case in the commit, unless I'm missing something.

I just did that to help with differentiating the cases. Maybe something was special about KASLR picking the wrong location that triggered the failure, etc.

Quoted 1 line

Maybe combine the prints as per the diff below?

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

That could work. If you're against the KASLR vs regular thing, I can respin the patch?

Happy to Ack it with that change, although I suppose it's ultimately up to Ard :)

No objections from me, but I prefer Will's version.

I took a look at this again... to report the failures as Will suggests, it would look like this:

drivers/firmware/efi/libstub/arm64-stub.c +3 -1
@@ -138,12 +138,14 @@ efi_status_t handle_kernel_image(efi_system_table_t *sys_table_arg,
        }
 
        if (status != EFI_SUCCESS) {
+               pr_efi_err(sys_table_arg, "allocate_pages() failed\n");
+
                *reserve_size = kernel_memsize + TEXT_OFFSET;
                status = efi_low_alloc(sys_table_arg, *reserve_size,
                                       MIN_KIMG_ALIGN, reserve_addr);
 
                if (status != EFI_SUCCESS) {
-                       pr_efi_err(sys_table_arg, "Failed to relocate kernel\n");
+                       pr_efi_err(sys_table_arg, "efi_low_alloc() failed\n");
                        *reserve_size = 0;
                        return status;
                }
Quoted 5 lines

My reasoning for putting the failure earlier is to differentiate which path was taken where the allocate_pages() failed: either regular or KASLR. If that's really not considered important here, I can send the above patch... Thoughts?

The first pr_efi_err() in the patch above complains about a condition that is not actually an error.

If you are interested in recording the path taken through this function, I have no objections to putting a normal pr_efi() print inside the KASLR block that shows that the physical placement of the kernel is being randomized. Then, we can keep only the second pr_efi_err() above to report the failure.