All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v4 0/4] Ensure 16 alignment of reserved memory in board_f.c
@ 2020-03-10  9:15 Patrick Delaunay
  2020-03-10  9:15 ` [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment Patrick Delaunay
                   ` (3 more replies)
  0 siblings, 4 replies; 9+ messages in thread
From: Patrick Delaunay @ 2020-03-10  9:15 UTC (permalink / raw)
  To: u-boot


Hi,

It is a V4 with only cosmetics udpate (s/insure/ensure)
for the V3 serie:
http://patchwork.ozlabs.org/project/uboot/list/?series=162918&state=*

V2 was:
http://patchwork.ozlabs.org/project/uboot/list/?series=154685&state=*

After the first correction, only for bootstage alignment,
I remove the stm32mp1 workaround in the 2nd patch.

The 4th patch is a complete solution to alignment
(proposed in comment 5 of
 http://patchwork.ozlabs.org/patch/1201452/#2327366)
I always align the reserved memory to 16 bytes with a new function
reserve_stack_aligned().

But this patch causes an issue on ARM 32 bits, as the relocated gd
pointer is not initialized with gd->new_gd as expected now in
reserve_global_data() but is hard-coded with:
  relocated gd = gd->bd - GD_SIZE
  {with GD_SIZE = sizeof(struct global_data)}

After the 4rd patch, this assumption in not more true as
sizeof(struct global_data) is not always 16 bytes aligned.

This issue is solved with the 3rd patch of the serie
  arm: set the relocated gd with gd->new_gd

Only tested on STM32MP157C-EV1 board (ARM32 architecture).


Changes in v4:
- replace insure by ensure in comment
- replace insure by ensure

Changes in v3:
- rename reserve_sp to reserve_stack_aligned

Changes in v2:
- import: [U-Boot,v3] board_f.c: Insure gd->new_bootstage alignment
- fix commit message s/bits/bytes/

Patrice Chotard (1):
  board_f.c: Ensure gd->new_bootstage alignment

Patrick Delaunay (3):
  Revert "stm32mp1: remove the imply BOOTSTAGE"
  arm: set the relocated gd with gd->new_gd
  board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory

 arch/arm/lib/crt0.S           |  3 +--
 arch/arm/mach-stm32mp/Kconfig |  2 ++
 common/board_f.c              | 27 ++++++++++++++++++---------
 3 files changed, 21 insertions(+), 11 deletions(-)

-- 
2.17.1

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment
  2020-03-10  9:15 [PATCH v4 0/4] Ensure 16 alignment of reserved memory in board_f.c Patrick Delaunay
@ 2020-03-10  9:15 ` Patrick Delaunay
  2020-04-17 21:08   ` Tom Rini
  2020-03-10  9:15 ` [PATCH v4 2/4] Revert "stm32mp1: remove the imply BOOTSTAGE" Patrick Delaunay
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 9+ messages in thread
From: Patrick Delaunay @ 2020-03-10  9:15 UTC (permalink / raw)
  To: u-boot

From: Patrice Chotard <patrice.chotard@st.com>

In reserve_bootstage(), in case size is odd, gd->new_bootstage
is not aligned. In bootstage_relocate(), the platform hangs when
getting access to data->record[i].name.
To avoid this issue, make gd->new_bootstage 16 byte aligned.

To ensure that new_bootstage is 16 byte aligned (at least needed for
x86_64 and ARMv8) and new_bootstage starts down to get enough space,
ALIGN_DOWN macro is used.

Fixes: ac9cd4805c8b ("bootstage: Correct relocation algorithm")

Signed-off-by: Patrice Chotard <patrice.chotard@st.com>
Reviewed-by: Vikas MANOCHA <vikas.manocha@st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@st.com>
Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
---

Changes in v4:
- replace insure by ensure in comment

Changes in v3: None
Changes in v2:
- import: [U-Boot,v3] board_f.c: Insure gd->new_bootstage alignment

 common/board_f.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/common/board_f.c b/common/board_f.c
index 82a164752a..0427b7b096 100644
--- a/common/board_f.c
+++ b/common/board_f.c
@@ -564,6 +564,11 @@ static int reserve_bootstage(void)
 	int size = bootstage_get_size();
 
 	gd->start_addr_sp -= size;
+	/*
+	 * Ensure that start_addr_sp is aligned down to reserve enough
+	 * space for new_bootstage
+	 */
+	gd->start_addr_sp = ALIGN_DOWN(gd->start_addr_sp, 16);
 	gd->new_bootstage = map_sysmem(gd->start_addr_sp, size);
 	debug("Reserving %#x Bytes for bootstage at: %08lx\n", size,
 	      gd->start_addr_sp);
-- 
2.17.1

^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v4 2/4] Revert "stm32mp1: remove the imply BOOTSTAGE"
  2020-03-10  9:15 [PATCH v4 0/4] Ensure 16 alignment of reserved memory in board_f.c Patrick Delaunay
  2020-03-10  9:15 ` [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment Patrick Delaunay
@ 2020-03-10  9:15 ` Patrick Delaunay
  2020-04-17 21:08   ` Tom Rini
  2020-03-10  9:15 ` [PATCH v4 3/4] arm: set the relocated gd with gd->new_gd Patrick Delaunay
  2020-03-10  9:15 ` [PATCH v4 4/4] board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory Patrick Delaunay
  3 siblings, 1 reply; 9+ messages in thread
From: Patrick Delaunay @ 2020-03-10  9:15 UTC (permalink / raw)
  To: u-boot

This reverts the workaround introduced by the
commit 16fec9b0bc1a ("stm32mp1: remove the imply BOOTSTAGE")
As the bootstage alignment issue is now solved.

Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
---

Changes in v4: None
Changes in v3: None
Changes in v2: None

 arch/arm/mach-stm32mp/Kconfig | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/arm/mach-stm32mp/Kconfig b/arch/arm/mach-stm32mp/Kconfig
index e4d621dee8..1489cbd5f3 100644
--- a/arch/arm/mach-stm32mp/Kconfig
+++ b/arch/arm/mach-stm32mp/Kconfig
@@ -62,7 +62,9 @@ config TARGET_ST_STM32MP15x
 	bool "STMicroelectronics STM32MP15x boards"
 	select STM32MP15x
 	imply BOOTCOUNT_LIMIT
+	imply BOOTSTAGE
 	imply CMD_BOOTCOUNT
+	imply CMD_BOOTSTAGE
 	imply CMD_CLS if CMD_BMP
 	imply DISABLE_CONSOLE
 	imply PRE_CONSOLE_BUFFER
-- 
2.17.1

^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v4 3/4] arm: set the relocated gd with gd->new_gd
  2020-03-10  9:15 [PATCH v4 0/4] Ensure 16 alignment of reserved memory in board_f.c Patrick Delaunay
  2020-03-10  9:15 ` [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment Patrick Delaunay
  2020-03-10  9:15 ` [PATCH v4 2/4] Revert "stm32mp1: remove the imply BOOTSTAGE" Patrick Delaunay
@ 2020-03-10  9:15 ` Patrick Delaunay
  2020-04-17 21:08   ` Tom Rini
  2020-03-10  9:15 ` [PATCH v4 4/4] board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory Patrick Delaunay
  3 siblings, 1 reply; 9+ messages in thread
From: Patrick Delaunay @ 2020-03-10  9:15 UTC (permalink / raw)
  To: u-boot

Simplify the arm relocation behavior and get gd directly form new_gd,
as it is already done in crt0_64.S:

	ldr	x18, [x18, #GD_NEW_GD]		/* x18 <- gd->new_gd */

This patch avoid assumption on new GD location (new GD is below bd -
with #GD_SIZE offset).

Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
---

Changes in v4: None
Changes in v3: None
Changes in v2: None

 arch/arm/lib/crt0.S | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/arch/arm/lib/crt0.S b/arch/arm/lib/crt0.S
index fb6c37cf51..df9dd83e40 100644
--- a/arch/arm/lib/crt0.S
+++ b/arch/arm/lib/crt0.S
@@ -127,8 +127,7 @@ ENTRY(_main)
 	ldr	r0, [r9, #GD_START_ADDR_SP]	/* sp = gd->start_addr_sp */
 	bic	r0, r0, #7	/* 8-byte alignment for ABI compliance */
 	mov	sp, r0
-	ldr	r9, [r9, #GD_BD]		/* r9 = gd->bd */
-	sub	r9, r9, #GD_SIZE		/* new GD is below bd */
+	ldr	r9, [r9, #GD_NEW_GD]		/* r9 <- gd->new_gd */
 
 	adr	lr, here
 	ldr	r0, [r9, #GD_RELOC_OFF]		/* r0 = gd->reloc_off */
-- 
2.17.1

^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v4 4/4] board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory
  2020-03-10  9:15 [PATCH v4 0/4] Ensure 16 alignment of reserved memory in board_f.c Patrick Delaunay
                   ` (2 preceding siblings ...)
  2020-03-10  9:15 ` [PATCH v4 3/4] arm: set the relocated gd with gd->new_gd Patrick Delaunay
@ 2020-03-10  9:15 ` Patrick Delaunay
  2020-04-17 21:08   ` Tom Rini
  3 siblings, 1 reply; 9+ messages in thread
From: Patrick Delaunay @ 2020-03-10  9:15 UTC (permalink / raw)
  To: u-boot

Add a function reserve_stack_aligned() to reserved memory with 16 bits
alignment after the stack pointer (gd->start_addr_sp) and use this new
function in board_f.c to reserve all the memory area (malloc, board, gd,
fdt, bootstage, stacks).

This 16 byte alignment is needed for cast on struct pointer
for the reserved memory, for example:
+ x86_64 ABI: https://reviews.llvm.org/D30049: 16 bytes
+ ARMv8 Instruction Set Overview: quad word, 16 bytes

An other alignment value could be needed for other architecture.

Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
---

Changes in v4:
- replace insure by ensure

Changes in v3:
- rename reserve_sp to reserve_stack_aligned

Changes in v2:
- fix commit message s/bits/bytes/

 common/board_f.c | 32 ++++++++++++++++++--------------
 1 file changed, 18 insertions(+), 14 deletions(-)

diff --git a/common/board_f.c b/common/board_f.c
index 0427b7b096..2ec5dbaa68 100644
--- a/common/board_f.c
+++ b/common/board_f.c
@@ -472,6 +472,17 @@ static int reserve_uboot(void)
 	return 0;
 }
 
+/*
+ * reserve after start_addr_sp the requested size and make the stack pointer
+ * 16-byte aligned, this alignment is needed for cast on the reserved memory
+ * ref = x86_64 ABI: https://reviews.llvm.org/D30049: 16 bytes
+ *     = ARMv8 Instruction Set Overview: quad word, 16 bytes
+ */
+static unsigned long reserve_stack_aligned(size_t size)
+{
+	return ALIGN_DOWN(gd->start_addr_sp - size, 16);
+}
+
 #ifdef CONFIG_SYS_NONCACHED_MEMORY
 static int reserve_noncached(void)
 {
@@ -497,7 +508,7 @@ static int reserve_noncached(void)
 /* reserve memory for malloc() area */
 static int reserve_malloc(void)
 {
-	gd->start_addr_sp = gd->start_addr_sp - TOTAL_MALLOC_LEN;
+	gd->start_addr_sp = reserve_stack_aligned(TOTAL_MALLOC_LEN);
 	debug("Reserving %dk for malloc() at: %08lx\n",
 	      TOTAL_MALLOC_LEN >> 10, gd->start_addr_sp);
 #ifdef CONFIG_SYS_NONCACHED_MEMORY
@@ -511,7 +522,7 @@ static int reserve_malloc(void)
 static int reserve_board(void)
 {
 	if (!gd->bd) {
-		gd->start_addr_sp -= sizeof(bd_t);
+		gd->start_addr_sp = reserve_stack_aligned(sizeof(bd_t));
 		gd->bd = (bd_t *)map_sysmem(gd->start_addr_sp, sizeof(bd_t));
 		memset(gd->bd, '\0', sizeof(bd_t));
 		debug("Reserving %zu Bytes for Board Info at: %08lx\n",
@@ -530,7 +541,7 @@ static int setup_machine(void)
 
 static int reserve_global_data(void)
 {
-	gd->start_addr_sp -= sizeof(gd_t);
+	gd->start_addr_sp = reserve_stack_aligned(sizeof(gd_t));
 	gd->new_gd = (gd_t *)map_sysmem(gd->start_addr_sp, sizeof(gd_t));
 	debug("Reserving %zu Bytes for Global Data at: %08lx\n",
 	      sizeof(gd_t), gd->start_addr_sp);
@@ -548,7 +559,7 @@ static int reserve_fdt(void)
 	if (gd->fdt_blob) {
 		gd->fdt_size = ALIGN(fdt_totalsize(gd->fdt_blob) + 0x1000, 32);
 
-		gd->start_addr_sp -= gd->fdt_size;
+		gd->start_addr_sp = reserve_stack_aligned(gd->fdt_size);
 		gd->new_fdt = map_sysmem(gd->start_addr_sp, gd->fdt_size);
 		debug("Reserving %lu Bytes for FDT at: %08lx\n",
 		      gd->fdt_size, gd->start_addr_sp);
@@ -563,12 +574,7 @@ static int reserve_bootstage(void)
 #ifdef CONFIG_BOOTSTAGE
 	int size = bootstage_get_size();
 
-	gd->start_addr_sp -= size;
-	/*
-	 * Ensure that start_addr_sp is aligned down to reserve enough
-	 * space for new_bootstage
-	 */
-	gd->start_addr_sp = ALIGN_DOWN(gd->start_addr_sp, 16);
+	gd->start_addr_sp = reserve_stack_aligned(size);
 	gd->new_bootstage = map_sysmem(gd->start_addr_sp, size);
 	debug("Reserving %#x Bytes for bootstage at: %08lx\n", size,
 	      gd->start_addr_sp);
@@ -585,8 +591,7 @@ __weak int arch_reserve_stacks(void)
 static int reserve_stacks(void)
 {
 	/* make stack pointer 16-byte aligned */
-	gd->start_addr_sp -= 16;
-	gd->start_addr_sp &= ~0xf;
+	gd->start_addr_sp = reserve_stack_aligned(16);
 
 	/*
 	 * let the architecture-specific code tailor gd->start_addr_sp and
@@ -598,8 +603,7 @@ static int reserve_stacks(void)
 static int reserve_bloblist(void)
 {
 #ifdef CONFIG_BLOBLIST
-	gd->start_addr_sp &= ~0xf;
-	gd->start_addr_sp -= CONFIG_BLOBLIST_SIZE;
+	gd->start_addr_sp = reserve_stack_aligned(CONFIG_BLOBLIST_SIZE);
 	gd->new_bloblist = map_sysmem(gd->start_addr_sp, CONFIG_BLOBLIST_SIZE);
 #endif
 
-- 
2.17.1

^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment
  2020-03-10  9:15 ` [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment Patrick Delaunay
@ 2020-04-17 21:08   ` Tom Rini
  0 siblings, 0 replies; 9+ messages in thread
From: Tom Rini @ 2020-04-17 21:08 UTC (permalink / raw)
  To: u-boot

On Tue, Mar 10, 2020 at 10:15:02AM +0100, Patrick Delaunay wrote:

> From: Patrice Chotard <patrice.chotard@st.com>
> 
> In reserve_bootstage(), in case size is odd, gd->new_bootstage
> is not aligned. In bootstage_relocate(), the platform hangs when
> getting access to data->record[i].name.
> To avoid this issue, make gd->new_bootstage 16 byte aligned.
> 
> To ensure that new_bootstage is 16 byte aligned (at least needed for
> x86_64 and ARMv8) and new_bootstage starts down to get enough space,
> ALIGN_DOWN macro is used.
> 
> Fixes: ac9cd4805c8b ("bootstage: Correct relocation algorithm")
> 
> Signed-off-by: Patrice Chotard <patrice.chotard@st.com>
> Reviewed-by: Vikas MANOCHA <vikas.manocha@st.com>
> Reviewed-by: Patrick Delaunay <patrick.delaunay@st.com>
> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
> Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>

Applied to u-boot/master, thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200417/6bf3b720/attachment.sig>

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v4 2/4] Revert "stm32mp1: remove the imply BOOTSTAGE"
  2020-03-10  9:15 ` [PATCH v4 2/4] Revert "stm32mp1: remove the imply BOOTSTAGE" Patrick Delaunay
@ 2020-04-17 21:08   ` Tom Rini
  0 siblings, 0 replies; 9+ messages in thread
From: Tom Rini @ 2020-04-17 21:08 UTC (permalink / raw)
  To: u-boot

On Tue, Mar 10, 2020 at 10:15:03AM +0100, Patrick Delaunay wrote:

> This reverts the workaround introduced by the
> commit 16fec9b0bc1a ("stm32mp1: remove the imply BOOTSTAGE")
> As the bootstage alignment issue is now solved.
> 
> Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>

Applied to u-boot/master, thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200417/06d07aaa/attachment.sig>

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v4 3/4] arm: set the relocated gd with gd->new_gd
  2020-03-10  9:15 ` [PATCH v4 3/4] arm: set the relocated gd with gd->new_gd Patrick Delaunay
@ 2020-04-17 21:08   ` Tom Rini
  0 siblings, 0 replies; 9+ messages in thread
From: Tom Rini @ 2020-04-17 21:08 UTC (permalink / raw)
  To: u-boot

On Tue, Mar 10, 2020 at 10:15:04AM +0100, Patrick Delaunay wrote:

> Simplify the arm relocation behavior and get gd directly form new_gd,
> as it is already done in crt0_64.S:
> 
> 	ldr	x18, [x18, #GD_NEW_GD]		/* x18 <- gd->new_gd */
> 
> This patch avoid assumption on new GD location (new GD is below bd -
> with #GD_SIZE offset).
> 
> Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>

Applied to u-boot/master, thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200417/f5709d45/attachment.sig>

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v4 4/4] board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory
  2020-03-10  9:15 ` [PATCH v4 4/4] board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory Patrick Delaunay
@ 2020-04-17 21:08   ` Tom Rini
  0 siblings, 0 replies; 9+ messages in thread
From: Tom Rini @ 2020-04-17 21:08 UTC (permalink / raw)
  To: u-boot

On Tue, Mar 10, 2020 at 10:15:05AM +0100, Patrick Delaunay wrote:

> Add a function reserve_stack_aligned() to reserved memory with 16 bits
> alignment after the stack pointer (gd->start_addr_sp) and use this new
> function in board_f.c to reserve all the memory area (malloc, board, gd,
> fdt, bootstage, stacks).
> 
> This 16 byte alignment is needed for cast on struct pointer
> for the reserved memory, for example:
> + x86_64 ABI: https://reviews.llvm.org/D30049: 16 bytes
> + ARMv8 Instruction Set Overview: quad word, 16 bytes
> 
> An other alignment value could be needed for other architecture.
> 
> Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
> Acked-by: Stephen Warren <swarren@nvidia.com>

Applied to u-boot/master, thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200417/1394297e/attachment.sig>

^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2020-04-17 21:08 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-03-10  9:15 [PATCH v4 0/4] Ensure 16 alignment of reserved memory in board_f.c Patrick Delaunay
2020-03-10  9:15 ` [PATCH v4 1/4] board_f.c: Ensure gd->new_bootstage alignment Patrick Delaunay
2020-04-17 21:08   ` Tom Rini
2020-03-10  9:15 ` [PATCH v4 2/4] Revert "stm32mp1: remove the imply BOOTSTAGE" Patrick Delaunay
2020-04-17 21:08   ` Tom Rini
2020-03-10  9:15 ` [PATCH v4 3/4] arm: set the relocated gd with gd->new_gd Patrick Delaunay
2020-04-17 21:08   ` Tom Rini
2020-03-10  9:15 ` [PATCH v4 4/4] board_f.c: Ensure 16 alignment of start_addr_sp and reserved memory Patrick Delaunay
2020-04-17 21:08   ` Tom Rini

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.