Skip to contentSkip to conversation

efi/arm64: fix fdt-related memory reservation

PATCH Superseded Mark Salter 4 participants 7 messages Sep 8, 2014

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

PATCH 1/1 efi/arm64: fix fdt-related memory reservation

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

This patch has the UEFI stub remove the memory reserve map entries from the FDT as it does with the memory nodes. This allows early_init_fdt_scan_reserved_mem() to be called unconditionally so that the other needed reservations are made.

---
 arch/arm64/mm/init.c               |  3 +--
 drivers/firmware/efi/libstub/fdt.c | 10 +++++++++-
 2 files changed, 10 insertions(+), 3 deletions(-)
arch/arm64/mm/init.c +1 -2
index 5472c24..a83061f 100644
@@ -149,8 +149,7 @@ void __init arm64_memblock_init(void)
 		memblock_reserve(__virt_to_phys(initrd_start), initrd_end - initrd_start);
 #endif
 
-	if (!efi_enabled(EFI_MEMMAP))
-		early_init_fdt_scan_reserved_mem();
+	early_init_fdt_scan_reserved_mem();
 
 	/* 4GB maximum for 32-bit only capable devices */
 	if (IS_ENABLED(CONFIG_ZONE_DMA))
drivers/firmware/efi/libstub/fdt.c +9 -1
index a56bb35..723c7da 100644
@@ -22,7 +22,7 @@ efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt,
 			unsigned long map_size, unsigned long desc_size,
 			u32 desc_ver)
 {
-	int node, prev;
+	int node, prev, i, num_rsv;
 	int status;
 	u32 fdt_val32;
 	u64 fdt_val64;
@@ -73,6 +73,14 @@ efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt,
 		prev = node;
 	}
 
+	/*
+	 * Delete all memory reserve map entries. When booting via UEFI,
+	 * kernel will use the UEFI memory map to find reserved regions.
+	 */
+	num_rsv = fdt_num_mem_rsv(fdt);
+	for (i = 0; i < num_rsv; i++)
+		fdt_del_mem_rsv(fdt, i);
+
 	node = fdt_subnode_offset(fdt, 0, "chosen");
 	if (node < 0) {
 		node = fdt_add_subnode(fdt, 0, "chosen");
Signature · 1 line
1.8.3.1
Acked-by on v1 1/1
#1 lore raw
Quoted 13 lines · Mark Salter

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

Quoted 6 lines

This patch has the UEFI stub remove the memory reserve map entries from the FDT as it does with the memory nodes. This allows early_init_fdt_scan_reserved_mem() to be called unconditionally so that the other needed reservations are made.

Interesting. I noticed the other day that my FDT got clobbered when accidentally booting with very little memory on QEMU , but I hadn't looked into it yet in more detail.

Matt: could you put this in your 'urgent' queue as well, please?

Cheers,
Ard.

Quoted 50 lines
---
 arch/arm64/mm/init.c               |  3 +--
 drivers/firmware/efi/libstub/fdt.c | 10 +++++++++-
 2 files changed, 10 insertions(+), 3 deletions(-)
arch/arm64/mm/init.c +1 -2
index 5472c24..a83061f 100644
@@ -149,8 +149,7 @@ void __init arm64_memblock_init(void)
                memblock_reserve(__virt_to_phys(initrd_start), initrd_end - initrd_start);
 #endif
 
-       if (!efi_enabled(EFI_MEMMAP))
-               early_init_fdt_scan_reserved_mem();
+       early_init_fdt_scan_reserved_mem();
 
        /* 4GB maximum for 32-bit only capable devices */
        if (IS_ENABLED(CONFIG_ZONE_DMA))
drivers/firmware/efi/libstub/fdt.c +9 -2
index a56bb35..723c7da 100644
@@ -22,7 +22,7 @@ efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt,
                        unsigned long map_size, unsigned long desc_size,
                        u32 desc_ver)
 {
-       int node, prev;
+       int node, prev, i, num_rsv;
        int status;
        u32 fdt_val32;
        u64 fdt_val64;
@@ -73,6 +73,14 @@ efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt,
                prev = node;
        }
 
+       /*
+        * Delete all memory reserve map entries. When booting via UEFI,
+        * kernel will use the UEFI memory map to find reserved regions.
+        */
+       num_rsv = fdt_num_mem_rsv(fdt);
+       for (i = 0; i < num_rsv; i++)
+               fdt_del_mem_rsv(fdt, i);
+
        node = fdt_subnode_offset(fdt, 0, "chosen");
        if (node < 0) {
                node = fdt_add_subnode(fdt, 0, "chosen");
--

1.8.3.1

reviewer Acked-by on v1 1/1
#2 lore raw
In reply to Mark Salter

Hi Mark,

Quoted 12 lines · Mark Salter

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

That doesn't sound like fun; apologies for allowing such brokenness through in the first place.

[...]

drivers/firmware/efi/libstub/fdt.c efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt, · line 82

Quoted 7 lines
drivers/firmware/efi/libstub/fdt.c +7 -0
+	/*
+	 * Delete all memory reserve map entries. When booting via UEFI,
+	 * kernel will use the UEFI memory map to find reserved regions.
+	 */
+	num_rsv = fdt_num_mem_rsv(fdt);
+	for (i = 0; i < num_rsv; i++)
+		fdt_del_mem_rsv(fdt, i);

I don't think that's right. Won't the memreserve entries shift down by one each time we call fdt_del_mem_rsv?

Shouldn't this be something like:

while (fdt_num_mem_rsv(fdt))

	fdt_del_mem_rsv(fdt, 0);

Or we could count downwards.

Otherwise, the general approach sounds sane to me, so with that bug fixed or disproven:

Given this is mostly in the EFI stub I expect that this will go via the EFI tree?

Mark.

Quoted 12 lines
drivers/firmware/efi/libstub/fdt.c +1 -0
+
 	node = fdt_subnode_offset(fdt, 0, "chosen");
 	if (node < 0) {
 		node = fdt_add_subnode(fdt, 0, "chosen");
Signature · 6 lines
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-efi" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#3 lore raw
Quoted 16 lines · Mark Rutland

Hi Mark,

Quoted 12 lines · Mark Salter

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

Quoted 2 lines

That doesn't sound like fun; apologies for allowing such brokenness through in the first place.

Heh. It was obvious that DT unflattening was broken, but bisecting didn't help much because I kept finding patches which when reverted made the problem go away even though they obviously weren't the cause.

Quoted 14 lines

[...]

drivers/firmware/efi/libstub/fdt.c efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt, · line 82

Quoted 7 lines
drivers/firmware/efi/libstub/fdt.c +7 -0
+	/*
+	 * Delete all memory reserve map entries. When booting via UEFI,
+	 * kernel will use the UEFI memory map to find reserved regions.
+	 */
+	num_rsv = fdt_num_mem_rsv(fdt);
+	for (i = 0; i < num_rsv; i++)
+		fdt_del_mem_rsv(fdt, i);

I don't think that's right. Won't the memreserve entries shift down by one each time we call fdt_del_mem_rsv?

Quoted 7 lines

Shouldn't this be something like:

while (fdt_num_mem_rsv(fdt))

	fdt_del_mem_rsv(fdt, 0);

Or we could count downwards.

Sigh. Yes, you are right. I only tested with one reserved region. I think counting down would be the way to go. I'll send a fixed patch shortly.

reviewer
#4 lore raw
Quoted 17 lines · Mark Salter
Quoted 16 lines · Mark Rutland

Hi Mark,

Quoted 12 lines · Mark Salter

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

Quoted 7 lines
Quoted 2 lines

That doesn't sound like fun; apologies for allowing such brokenness through in the first place.

Heh. It was obvious that DT unflattening was broken, but bisecting didn't help much because I kept finding patches which when reverted made the problem go away even though they obviously weren't the cause.

Yeah, those types of bugs are never fun. I recall a similar situation with the __INIT annotation in the versatile pen code.

Quoted 18 lines
Quoted 18 lines

[...]

drivers/firmware/efi/libstub/fdt.c efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt, · line 82

Quoted 7 lines
drivers/firmware/efi/libstub/fdt.c +7 -0
+	/*
+	 * Delete all memory reserve map entries. When booting via UEFI,
+	 * kernel will use the UEFI memory map to find reserved regions.
+	 */
+	num_rsv = fdt_num_mem_rsv(fdt);
+	for (i = 0; i < num_rsv; i++)
+		fdt_del_mem_rsv(fdt, i);

I don't think that's right. Won't the memreserve entries shift down by one each time we call fdt_del_mem_rsv?

Shouldn't this be something like:

while (fdt_num_mem_rsv(fdt))

	fdt_del_mem_rsv(fdt, 0);
Quoted 6 lines
Quoted 2 lines

Or we could count downwards.

Sigh. Yes, you are right. I only tested with one reserved region. I think counting down would be the way to go. I'll send a fixed patch shortly.

Ok, that sounds fine by me; make sure you add my ack :)

Thanks,
Mark.

#5 lore raw

PATCH v2 1/1 efi/arm64: fix fdt-related memory reservation

In reply to Mark Salter

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

This patch has the UEFI stub remove the memory reserve map entries from the FDT as it does with the memory nodes. This allows early_init_fdt_scan_reserved_mem() to be called unconditionally so that the other needed reservations are made.

Change from v1:
* Acks added
* Fixed loop logic to remove reserved regions

---
 arch/arm64/mm/init.c               |  3 +--
 drivers/firmware/efi/libstub/fdt.c | 10 +++++++++-
 2 files changed, 10 insertions(+), 3 deletions(-)
arch/arm64/mm/init.c +1 -2
index 5472c24..a83061f 100644
@@ -149,8 +149,7 @@ void __init arm64_memblock_init(void)
 		memblock_reserve(__virt_to_phys(initrd_start), initrd_end - initrd_start);
 #endif
 
-	if (!efi_enabled(EFI_MEMMAP))
-		early_init_fdt_scan_reserved_mem();
+	early_init_fdt_scan_reserved_mem();
 
 	/* 4GB maximum for 32-bit only capable devices */
 	if (IS_ENABLED(CONFIG_ZONE_DMA))
drivers/firmware/efi/libstub/fdt.c +9 -1
index a56bb35..c846a96 100644
@@ -22,7 +22,7 @@ efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt,
 			unsigned long map_size, unsigned long desc_size,
 			u32 desc_ver)
 {
-	int node, prev;
+	int node, prev, num_rsv;
 	int status;
 	u32 fdt_val32;
 	u64 fdt_val64;
@@ -73,6 +73,14 @@ efi_status_t update_fdt(efi_system_table_t *sys_table, void *orig_fdt,
 		prev = node;
 	}
 
+	/*
+	 * Delete all memory reserve map entries. When booting via UEFI,
+	 * kernel will use the UEFI memory map to find reserved regions.
+	 */
+	num_rsv = fdt_num_mem_rsv(fdt);
+	while (num_rsv-- > 0)
+		fdt_del_mem_rsv(fdt, num_rsv);
+
 	node = fdt_subnode_offset(fdt, 0, "chosen");
 	if (node < 0) {
 		node = fdt_add_subnode(fdt, 0, "chosen");
Signature · 1 line
1.8.3.1
#6 lore raw
Quoted 22 lines · Mark Salter

Commit 86c8b27a01cf:
"arm64: ignore DT memreserve entries when booting in UEFI mode

prevents early_init_fdt_scan_reserved_mem() from being called for arm64 kernels booting via UEFI. This was done because the kernel will use the UEFI memory map to determine reserved memory regions. That approach has problems in that early_init_fdt_scan_reserved_mem() also reserves the FDT itself and any node-specific reserved memory. By chance of some kernel configs, the FDT may be overwritten before it can be unflattened and the kernel will fail to boot. More subtle problems will result if the FDT has node specific reserved memory which is not really reserved.

This patch has the UEFI stub remove the memory reserve map entries from the FDT as it does with the memory nodes. This allows early_init_fdt_scan_reserved_mem() to be called unconditionally so that the other needed reservations are made.

Change from v1:
* Acks added
* Fixed loop logic to remove reserved regions

Quoted 7 lines
---
 arch/arm64/mm/init.c               |  3 +--
 drivers/firmware/efi/libstub/fdt.c | 10 +++++++++-
 2 files changed, 10 insertions(+), 3 deletions(-)

Thanks Mark, queued up in 'urgent'.

Signature · 1 line
Matt Fleming, Intel Open Source Technology Center