Skip to contentSkip to conversation

remoteproc: imx_rproc: merge ITCM and DTCM regions

PATCH In review Hiago De Franco 4 participants 9 messages Jul 3, 2025 – Jul 10, 2025

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

PATCH 1/1 remoteproc: imx_rproc: merge ITCM and DTCM regions

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -4
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5
maintainer
#1 lore raw

Good morning,

Quoted 51 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -2
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = { · line 184

Quoted 5 lines
drivers/remoteproc/imx_rproc.c +0 -2
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },

In commit 8749919defb8 "dev addr" and "sys addr" were both contiguous, but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5
#2 lore raw

Hi Mathieu,

Quoted 55 lines · Mathieu Poirier

Good morning,

Quoted 52 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -2
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
Quoted 8 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = { · line 184

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },

In commit 8749919defb8 "dev addr" and "sys addr" were both contiguous, but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for DTCM entry.

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM, instead of going to OCRAM or DDR.

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5
#3 lore raw
Quoted 77 lines

Subject: Re: [PATCH] remoteproc: imx_rproc: merge ITCM and DTCM regions

Hi Mathieu,

Quoted 6 lines · Mathieu Poirier

Good morning,

Quoted 3 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region

to

Quoted 5 lines
Quoted 5 lines

prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc remoteproc
remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len =

0x1fcb0

Quoted 1 line
Quoted 1 line

remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0

remoteproc

Quoted 20 lines
Quoted 20 lines

remoteproc0: Failed to load program segments: -22 remoteproc
remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Quoted 8 lines
Quoted 8 lines

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
(quoted diff) +0 -0

b/drivers/remoteproc/imx_rproc.c index

74299af1d7f1..bbf089ef48ee

Quoted 10 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att · line 171

Quoted 10 lines

100644

drivers/remoteproc/imx_rproc.c +0 -2
@@ -166,8 +166,8 @@ static const struct imx_rproc_att
 imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN |

ATT_IOMEM },

Quoted 2 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att · line 171

Quoted 2 lines
drivers/remoteproc/imx_rproc.c +2 -0
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN |

ATT_IOMEM },

Quoted 4 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att

imx_rproc_att_imx8mn[] = {

Quoted 5 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {

Quoted 5 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN |

ATT_IOMEM },

Quoted 2 lines

In commit 8749919defb8 "dev addr" and "sys addr" were both

Quoted 9 lines

contiguous,

Quoted 2 lines

but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for DTCM entry.

The ITCM in M7 starts from 0, DTCM in M7 starts from 0x20000000. The ITCM/DTCM is continuous from A53 perspective.

You let linux to load continuous text/data, but if the dev addr start from ITCM_START+128KB is not reserved from Reference manual.

How could M7 still work proper if M7 accessing [0x20000, x] dev_addr?

You need update your linker script to separate text into different sections.

Regards,
Peng.

Quoted 2 lines

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM,

instead of going to OCRAM

Quoted 15 lines

or DDR.

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -1
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
--

2.39.5

#4 lore raw
Quoted 88 lines

Subject: RE: [PATCH] remoteproc: imx_rproc: merge ITCM and DTCM regions

Quoted 77 lines

Subject: Re: [PATCH] remoteproc: imx_rproc: merge ITCM and DTCM regions

Hi Mathieu,

Quoted 6 lines · Mathieu Poirier

Good morning,

Quoted 3 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region

to

Quoted 5 lines
Quoted 5 lines

prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc remoteproc
remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len =

0x1fcb0

Quoted 1 line
Quoted 1 line

remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0

remoteproc

Quoted 20 lines
Quoted 20 lines

remoteproc0: Failed to load program segments: -22 remoteproc
remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Quoted 8 lines
Quoted 8 lines

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
(quoted diff) +0 -0

b/drivers/remoteproc/imx_rproc.c index

74299af1d7f1..bbf089ef48ee

Quoted 10 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att · line 171

Quoted 10 lines

100644

drivers/remoteproc/imx_rproc.c +0 -2
@@ -166,8 +166,8 @@ static const struct imx_rproc_att
 imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN |

ATT_IOMEM },

Quoted 2 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att · line 171

Quoted 2 lines
drivers/remoteproc/imx_rproc.c +2 -0
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN |

ATT_IOMEM },

Quoted 4 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att

imx_rproc_att_imx8mn[] = {

Quoted 5 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {

Quoted 5 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN |

ATT_IOMEM },

Quoted 2 lines

In commit 8749919defb8 "dev addr" and "sys addr" were both

Quoted 8 lines

contiguous,

Quoted 2 lines

but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for

Quoted 8 lines

DTCM entry.

Quoted 1 line

The ITCM in M7 starts from 0, DTCM in M7 starts from 0x20000000. The ITCM/DTCM is continuous from A53 perspective.

You let linux to load continuous text/data, but if the dev addr start from ITCM_START+128KB is not reserved from Reference manual.

Typo: ITCM_START + 128KB is reserved.

Quoted 30 lines

How could M7 still work proper if M7 accessing [0x20000, x] dev_addr?

You need update your linker script to separate text into different sections.

Regards,
Peng.

Quoted 2 lines

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM,

instead of going to OCRAM

Quoted 15 lines

or DDR.

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -1
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
--

2.39.5

maintainer
#5 lore raw
In reply to Hiago De Franco
Quoted 67 lines · Hiago De Franco

Hi Mathieu,

Quoted 55 lines · Mathieu Poirier

Good morning,

Quoted 52 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -2
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
Quoted 8 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = { · line 184

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },

In commit 8749919defb8 "dev addr" and "sys addr" were both contiguous, but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

Quoted 5 lines

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for DTCM entry.

Correct. Older firmware would still expect DTCM at 0x20000000.

Quoted 4 lines

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM, instead of going to OCRAM or DDR.

To me the best way to proceed is understand why using the current mapping is a problem. The changelog describes a failure condition when dealing with large sections but does not indicate _why_ that is happening. I think that's what needs to be fixed rather than trying to move mappings around.

Quoted 13 lines

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5
#6 lore raw

Hi Peng, Mathieu,

Quoted 82 lines · Mathieu Poirier
Quoted 67 lines · Hiago De Franco

Hi Mathieu,

Quoted 55 lines · Mathieu Poirier

Good morning,

Quoted 52 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -2
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
Quoted 8 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = { · line 184

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },

In commit 8749919defb8 "dev addr" and "sys addr" were both contiguous, but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

Quoted 5 lines

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for DTCM entry.

Correct. Older firmware would still expect DTCM at 0x20000000.

Quoted 4 lines

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM, instead of going to OCRAM or DDR.

Quoted 4 lines

To me the best way to proceed is understand why using the current mapping is a problem. The changelog describes a failure condition when dealing with large sections but does not indicate _why_ that is happening. I think that's what needs to be fixed rather than trying to move mappings around.

Thanks for the information you both provided, sorry for the noise, so please Mathieu ignore this patch. I will work around this in a different way.

By the way, Peng, I noticed the DDR linker from MCUXpresso does not work if the firmware is larger than 128KB, since the .data is placed right after .text and loaded later to DDR. The imx_rproc driver should instead have a way to do the other way around: starting from the firwmare inside DDR, we could set PC and stack from M7 to point to DDR and start the execution. Probably will be slower, but it would make the DDR case possible.

Correct me if I am wrong, but as my current understanding the DDR linker is broken without this change to the driver. Anyway, maybe something for a future patch.

Thanks,
Hiago.

Quoted 14 lines
Quoted 14 lines

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5
#7 lore raw
Quoted 94 lines · Hiago De Franco

Hi Peng, Mathieu,

Quoted 82 lines · Mathieu Poirier
Quoted 67 lines · Hiago De Franco

Hi Mathieu,

Quoted 55 lines · Mathieu Poirier

Good morning,

Quoted 52 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -2
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
Quoted 8 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = { · line 184

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },

In commit 8749919defb8 "dev addr" and "sys addr" were both contiguous, but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

Quoted 5 lines

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for DTCM entry.

Correct. Older firmware would still expect DTCM at 0x20000000.

Quoted 4 lines

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM, instead of going to OCRAM or DDR.

Quoted 4 lines

To me the best way to proceed is understand why using the current mapping is a problem. The changelog describes a failure condition when dealing with large sections but does not indicate _why_ that is happening. I think that's what needs to be fixed rather than trying to move mappings around.

Thanks for the information you both provided, sorry for the noise, so please Mathieu ignore this patch. I will work around this in a different way.

Quoted 7 lines

By the way, Peng, I noticed the DDR linker from MCUXpresso does not work if the firmware is larger than 128KB, since the .data is placed right after .text and loaded later to DDR. The imx_rproc driver should instead have a way to do the other way around: starting from the firwmare inside DDR, we could set PC and stack from M7 to point to DDR and start the execution. Probably will be slower, but it would make the DDR case possible.

I am not aware of the size limitation if image is built to run in DDR. It maybe MCUXpresso team just reuse the linker script for TCM and only update the link address.

You could update the linker script to build larger image.

In final elf, .data is put just after .text, but the related section loading address should be specified as my understanding. See below, .data is at 0x20000000 for M7.

xx-gcc -S imx8mn_m7_TCM_rpmsg_lite_str_echo_rtos.elf There are 20 section headers, starting at offset 0xc998:

Section Headers:

  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [ 0]                   NULL            00000000 000000 000000 00      0   0  0
  [ 1] .interrupts       PROGBITS        00000000 001000 000240 00   A  0   0  4
  [ 2] .resource_table   PROGBITS        00000240 001240 000058 00   A  0   0  1
  [ 3] .text             PROGBITS        000002a0 0012a0 0046b0 00  AX  0   0 16
  [ 4] .ARM              ARM_EXIDX       00004950 005950 000008 00  AL  3   0  4
  [ 5] .init_array       INIT_ARRAY      00004958 005958 000004 04  WA  0   0  4
  [ 6] .fini_array       FINI_ARRAY      0000495c 00595c 000004 04  WA  0   0  4
  [ 7] .data             PROGBITS        20000000 006000 00000c 00  WA  0   0  4
  [ 8] .ncache.init      PROGBITS        80000000 00600c 000000 00   W  0   0  1
  [ 9] .ncache           PROGBITS        80000000 00600c 000000 00   W  0   0  1
  [10] .bss              NOBITS          2000000c 00600c 00a4ac 00  WA  0   0  4
  [11] .heap             NOBITS          2000a4b8 00600c 000400 00  WA  0   0  1
  [12] .stack            NOBITS          2000a8b8 00600c 000400 00  WA  0   0  1
  [13] .ARM.attributes   ARM_ATTRIBUTES  00000000 00600c 000030 00      0   0  1
  [14] .debug_line_str   PROGBITS        00000000 00603c 0002b1 01  MS  0   0  1
  [15] .comment          PROGBITS        00000000 0062ed 000055 01  MS  0   0  1
  [16] .debug_frame      PROGBITS        00000000 006344 000260 00      0   0  4
  [17] .symtab           SYMTAB          00000000 0065a4 003cd0 10     18 586  4
  [18] .strtab           STRTAB          00000000 00a274 002664 00      0   0  1
  [19] .shstrtab         STRTAB          00000000 00c8d8 0000bd 00      0   0  1
Quoted 4 lines

Correct me if I am wrong, but as my current understanding the DDR linker is broken without this change to the driver. Anyway, maybe something for a future patch.

If you wanna image in DDR, you could specifiy the address of data section in your linker script.

But for support ddr elf file, you need patches as below:
https://lore.kernel.org/linux-arm-kernel/CAEnQRZC5t=qmo+OJLW+dqZg4gH9cAN=paWDSGbrJb2AvkKBqxg@mail.gmail.com/T/#ec54c42b70416b002936a643b44b79661dd2a8483
This patchset was rejected, because we need to get stack/pc from .interrupts section and store to ITCM.

Latest NXP m7 demo has included a new section for stack/pc, but this will only be public in 2025 Q3 release, for pre-2025-Q3 releases, still need the upper patchset.

Regards,
Peng

Quoted 18 lines

Thanks,
Hiago.

Quoted 14 lines
Quoted 14 lines

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5
#8 lore raw
Quoted 147 lines · Peng Fan
Quoted 94 lines · Hiago De Franco

Hi Peng, Mathieu,

Quoted 82 lines · Mathieu Poirier
Quoted 67 lines · Hiago De Franco

Hi Mathieu,

Quoted 55 lines · Mathieu Poirier

Good morning,

Quoted 52 lines · Hiago De Franco

From: Hiago De Franco <[email protected]>

Merge the contiguous ITCM and DTCM regions into a single region to prevent failures when loading ELF files with large sections:

remoteproc remoteproc0: powering up imx-rproc
remoteproc remoteproc0: Booting fw image rproc-imx-rproc-fw, size 151824 imx-rproc imx8mp-cm7: Translation failed: da = 0x1f48 len = 0x1fcb0 remoteproc remoteproc0: bad phdr da 0x1f48 mem 0x1fcb0 remoteproc remoteproc0: Failed to load program segments: -22 remoteproc remoteproc0: Boot failed: -22

This approach is the same as commit 8749919defb8 ("remoteproc:
imx_rproc: Merge TCML/U").

---
Hi,

The ELF I tested had the following data section:

Memory region Used Size Region Size %age Used
m_interrupts: 680 B 1 KB 66.41%
m_text: 6984 B 127 KB 5.37%
m_data: 130224 B 128 KB 99.35%
m_data2: 0 GB 16 MB 0.00%
[100%] Built target hello_world_cm7.elf

Which triggered the error. After this patch, remoteproc was able to boot and work fine. Thanks!

---
 drivers/remoteproc/imx_rproc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/remoteproc/imx_rproc.c +2 -2
index 74299af1d7f1..bbf089ef48ee 100644
@@ -166,8 +166,8 @@ static const struct imx_rproc_att imx_rproc_att_imx8qxp[] = {
 
 static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	/* dev addr , sys addr  , size	    , flags */
-	/* ITCM   */
-	{ 0x00000000, 0x007E0000, 0x00020000, ATT_OWN | ATT_IOMEM },
+	/* D/ITCM */
+	{ 0x00000000, 0x007E0000, 0x00040000, ATT_OWN | ATT_IOMEM },
 	/* OCRAM_S */
 	{ 0x00180000, 0x00180000, 0x00009000, 0 },
 	/* OCRAM */
@@ -180,8 +180,6 @@ static const struct imx_rproc_att imx_rproc_att_imx8mn[] = {
 	{ 0x08000000, 0x08000000, 0x08000000, 0 },
Quoted 8 lines

drivers/remoteproc/imx_rproc.c static const struct imx_rproc_att imx_rproc_att_imx8mn[] = { · line 184

Quoted 4 lines
drivers/remoteproc/imx_rproc.c +0 -2
 	/* DDR (Code) - alias */
 	{ 0x10000000, 0x40000000, 0x0FFE0000, 0 },
-	/* DTCM */
-	{ 0x20000000, 0x00800000, 0x00020000, ATT_OWN | ATT_IOMEM },

In commit 8749919defb8 "dev addr" and "sys addr" were both contiguous, but in this patch "dev addr" is not. How will this work with new kernel that use old FW images? Am I missing something?

Quoted 5 lines

No, you are correct, I think the use case I tested was not good enough.

If I understand correctly, this will break older firmware expecting .data at 0x20000000 because dev_addr is no longer mapped for DTCM entry.

Correct. Older firmware would still expect DTCM at 0x20000000.

Quoted 4 lines

Do you think it is possible (or reccomend) another approach to fix this issue? In this case to keep using the TCM, instead of going to OCRAM or DDR.

Quoted 4 lines

To me the best way to proceed is understand why using the current mapping is a problem. The changelog describes a failure condition when dealing with large sections but does not indicate _why_ that is happening. I think that's what needs to be fixed rather than trying to move mappings around.

Thanks for the information you both provided, sorry for the noise, so please Mathieu ignore this patch. I will work around this in a different way.

Quoted 7 lines

By the way, Peng, I noticed the DDR linker from MCUXpresso does not work if the firmware is larger than 128KB, since the .data is placed right after .text and loaded later to DDR. The imx_rproc driver should instead have a way to do the other way around: starting from the firwmare inside DDR, we could set PC and stack from M7 to point to DDR and start the execution. Probably will be slower, but it would make the DDR case possible.

I am not aware of the size limitation if image is built to run in DDR. It maybe MCUXpresso team just reuse the linker script for TCM and only update the link address.

You could update the linker script to build larger image.

In final elf, .data is put just after .text, but the related section loading address should be specified as my understanding. See below, .data is at 0x20000000 for M7.

xx-gcc -S imx8mn_m7_TCM_rpmsg_lite_str_echo_rtos.elf There are 20 section headers, starting at offset 0xc998:

Section Headers:

  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [ 0]                   NULL            00000000 000000 000000 00      0   0  0
  [ 1] .interrupts       PROGBITS        00000000 001000 000240 00   A  0   0  4
  [ 2] .resource_table   PROGBITS        00000240 001240 000058 00   A  0   0  1
  [ 3] .text             PROGBITS        000002a0 0012a0 0046b0 00  AX  0   0 16
  [ 4] .ARM              ARM_EXIDX       00004950 005950 000008 00  AL  3   0  4
  [ 5] .init_array       INIT_ARRAY      00004958 005958 000004 04  WA  0   0  4
  [ 6] .fini_array       FINI_ARRAY      0000495c 00595c 000004 04  WA  0   0  4
  [ 7] .data             PROGBITS        20000000 006000 00000c 00  WA  0   0  4
  [ 8] .ncache.init      PROGBITS        80000000 00600c 000000 00   W  0   0  1
  [ 9] .ncache           PROGBITS        80000000 00600c 000000 00   W  0   0  1
  [10] .bss              NOBITS          2000000c 00600c 00a4ac 00  WA  0   0  4
  [11] .heap             NOBITS          2000a4b8 00600c 000400 00  WA  0   0  1
  [12] .stack            NOBITS          2000a8b8 00600c 000400 00  WA  0   0  1
  [13] .ARM.attributes   ARM_ATTRIBUTES  00000000 00600c 000030 00      0   0  1
  [14] .debug_line_str   PROGBITS        00000000 00603c 0002b1 01  MS  0   0  1
  [15] .comment          PROGBITS        00000000 0062ed 000055 01  MS  0   0  1
  [16] .debug_frame      PROGBITS        00000000 006344 000260 00      0   0  4
  [17] .symtab           SYMTAB          00000000 0065a4 003cd0 10     18 586  4
  [18] .strtab           STRTAB          00000000 00a274 002664 00      0   0  1
  [19] .shstrtab         STRTAB          00000000 00c8d8 0000bd 00      0   0  1
Quoted 4 lines

Correct me if I am wrong, but as my current understanding the DDR linker is broken without this change to the driver. Anyway, maybe something for a future patch.

If you wanna image in DDR, you could specifiy the address of data section in your linker script.

Quoted 8 lines

But for support ddr elf file, you need patches as below:
https://lore.kernel.org/linux-arm-kernel/CAEnQRZC5t=qmo+OJLW+dqZg4gH9cAN=paWDSGbrJb2AvkKBqxg@mail.gmail.com/T/#ec54c42b70416b002936a643b44b79661dd2a8483
This patchset was rejected, because we need to get stack/pc from .interrupts section and store to ITCM.

Latest NXP m7 demo has included a new section for stack/pc, but this will only be public in 2025 Q3 release, for pre-2025-Q3 releases, still need the upper patchset.

Great, thanks for all the information Peng. I will inform my team here.

Best regards,
Hiago.

Quoted 22 lines

Regards,
Peng

Quoted 18 lines

Thanks,
Hiago.

Quoted 14 lines
Quoted 14 lines

Thanks,
Hiago.

Quoted 10 lines

Thanks,
Mathieu

Quoted 6 lines
drivers/remoteproc/imx_rproc.c +0 -0
 	/* OCRAM_S - alias */
 	{ 0x20180000, 0x00180000, 0x00008000, ATT_OWN },
 	/* OCRAM */
Signature · 1 line
2.39.5