Skip to contentSkip to conversation

tty: amba-pl011: Use 32-bit accesses for SBSA UART

PATCH In review Christopher Covington 6 participants 12 messages Mar 8, 2016 – Apr 1, 2016

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tty: amba-pl011: Use 32-bit accesses for SBSA UART

From: Christopher Covington <[email protected]>

Version 2 of the Server Base System Architecture (SBSA) describes the Generic UART registers as 32 bits wide. At least one implementation, found of the Qualcomm Technologies QDF2432, only supports 32 bit accesses. While other implementations may also support smaller sized accesses, simply use 32 bit accesses all the time for the SBSA UART for simple, broad, compatibility.

---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +1 -0
index c0da0cc..ffb5eb8 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,
Signature · 3 lines
Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
Tested-by on v1 1/1
#1 lore raw
Quoted 11 lines · Christopher Covington

From: Christopher Covington <[email protected]>

Version 2 of the Server Base System Architecture (SBSA) describes the Generic UART registers as 32 bits wide. At least one implementation, found of the Qualcomm Technologies QDF2432, only supports 32 bit accesses. While other implementations may also support smaller sized accesses, simply use 32 bit accesses all the time for the SBSA UART for simple, broad, compatibility.

---

I can boot the v4.5-rc7 on my QDF2432 platform with this patch, and could not otherwise.

#2 lore raw

PATCH 1/1 tty: amba-pl011: Use 32-bit accesses for SBSA UART

In reply to Christopher Covington

Version 2 of the Server Base System Architecture (SBSAv2) describes the Generic UART registers as 32 bits wide. At least one implementation, found on the Qualcomm Technologies QDF2432, only supports 32 bit accesses. SBSAv3, which describes supported access sizes in greater detail, explicitly requires support for both 16 and 32 bit accesses to all registers (and 8 bit accesses to some but not all). Therefore, for broad compatibility, simply use 32 bit accessors for the SBSA UART.

---
Changes new in v2:
* Fixed from address
* Elaborated on forward (SBSAv3) compatibility in commit message
* Included Mark Langsdorf's Tested-by, which now covers:

    QDF2432
    Seattle
    X-Gene 1
---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +1 -0
index c0da0cc..ffb5eb8 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,
Signature · 3 lines
Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#3 lore raw

Hi Christopher,

Quoted 7 lines · Christopher Covington

Version 2 of the Server Base System Architecture (SBSAv2) describes the Generic UART registers as 32 bits wide. At least one implementation, found on the Qualcomm Technologies QDF2432, only supports 32 bit accesses. SBSAv3, which describes supported access sizes in greater detail, explicitly requires support for both 16 and 32 bit accesses to all registers (and 8 bit accesses to some but not all). Therefore, for broad compatibility, simply use 32 bit accessors for the SBSA UART.

So this eliminates the need to configure SBSA port via ACPI, correct? Thus, Aleksey can drop his "serial: pl011: use SPCR to setup 32-bit access"?

Quoted 27 lines

---
Changes new in v2:
* Fixed from address
* Elaborated on forward (SBSAv3) compatibility in commit message
* Included Mark Langsdorf's Tested-by, which now covers:

    QDF2432
    Seattle
    X-Gene 1
---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +1 -0
index c0da0cc..ffb5eb8 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,
#4 lore raw
Quoted 11 lines · Peter Hurley

Hi Christopher,

Quoted 1 line · Christopher Covington

Version 2 of the Server Base System Architecture (SBSAv2) describes

the

Quoted 1 line

Generic UART registers as 32 bits wide. At least one implementation,

found

Quoted 4 lines

on the Qualcomm Technologies QDF2432, only supports 32 bit accesses. SBSAv3, which describes supported access sizes in greater detail, explicitly requires support for both 16 and 32 bit accesses to all registers (and 8 bit accesses to some but not all). Therefore, for

Quoted 6 lines

broad

Quoted 1 line

compatibility, simply use 32 bit accessors for the SBSA UART.

So this eliminates the need to configure SBSA port via ACPI, correct? Thus, Aleksey can drop his "serial: pl011: use SPCR to setup 32-bit access"?

Yes.

Thanks,
Christopher Covington

Signature · 5 lines
Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

Sent from my Snapdragon powered Android device with K-9 Mail. Please excuse my brevity.
Tested-by Acked-by on v1 1/1
#5 lore raw
In reply to Christopher Covington

Hi Christopher,

Quoted 10 lines · Christopher Covington

Version 2 of the Server Base System Architecture (SBSAv2) describes the Generic UART registers as 32 bits wide. At least one implementation, found on the Qualcomm Technologies QDF2432, only supports 32 bit accesses. SBSAv3, which describes supported access sizes in greater detail, explicitly requires support for both 16 and 32 bit accesses to all registers (and 8 bit accesses to some but not all). Therefore, for broad compatibility, simply use 32 bit accessors for the SBSA UART.

So I gave this a try on a Juno and a Midway. Both have a normal PL011, but I changed the DT to advertise an SBSA UART instead.
This worked fine with the 32bit accessors.
Also according to some research on the hardware size at least the current ARM PL011 implementation are totally fine with 32-bit (as well as 16-bit) accesses.
There is some reluctance about whether this is true for _every_ older PL011 implementation, but they are out of scope here, as we are talking about the SBSA only.
So:

You can add Juno and Midway to the list of tested systems.

Cheers,
Andre.

Quoted 31 lines

---
Changes new in v2:
* Fixed from address
* Elaborated on forward (SBSAv3) compatibility in commit message
* Included Mark Langsdorf's Tested-by, which now covers:

    QDF2432
    Seattle
    X-Gene 1
---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +1 -0
index c0da0cc..ffb5eb8 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,
#6 lore raw

Hi Greg,

Quoted 25 lines · Andre Przywara

Hi Christopher,

Quoted 10 lines · Christopher Covington

Version 2 of the Server Base System Architecture (SBSAv2) describes the Generic UART registers as 32 bits wide. At least one implementation, found on the Qualcomm Technologies QDF2432, only supports 32 bit accesses. SBSAv3, which describes supported access sizes in greater detail, explicitly requires support for both 16 and 32 bit accesses to all registers (and 8 bit accesses to some but not all). Therefore, for broad compatibility, simply use 32 bit accessors for the SBSA UART.

So I gave this a try on a Juno and a Midway. Both have a normal PL011, but I changed the DT to advertise an SBSA UART instead.
This worked fine with the 32bit accessors.
Also according to some research on the hardware size at least the current ARM PL011 implementation are totally fine with 32-bit (as well as 16-bit) accesses.
There is some reluctance about whether this is true for _every_ older PL011 implementation, but they are out of scope here, as we are talking about the SBSA only.
So:

Quoted 6 lines

You can add Juno and Midway to the list of tested systems.

Quoted 1 line

Changes new in v2:

Apologies for omitting the v2 prefix in the second version of the patch that I sent out.

Quoted 17 lines
Quoted 17 lines

* Fixed from address
* Elaborated on forward (SBSAv3) compatibility in commit message
* Included Mark Langsdorf's Tested-by, which now covers:

    QDF2432
    Seattle
    X-Gene 1
---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +0 -0
index c0da0cc..ffb5eb8 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
Quoted 7 lines

drivers/tty/serial/amba-pl011.c static struct vendor_data vendor_arm = { · line 127

Quoted 7 lines
drivers/tty/serial/amba-pl011.c +1 -0
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,

Do you consider this patch suitable to be included in a 4.6 release candidate? It fixes an issue running this driver on certain hardware, and with gracious assistance we've performed due diligence to check that it does not adversely affect the driver running on other hardware. Would it be useful to send a v3 collecting the acks and tested-bys?

Thanks,
Christopher Covington

Signature · 3 lines
Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#7 lore raw
Quoted 55 lines · Christopher Covington

Hi Greg,

Quoted 25 lines · Andre Przywara

Hi Christopher,

Quoted 10 lines · Christopher Covington

Version 2 of the Server Base System Architecture (SBSAv2) describes the Generic UART registers as 32 bits wide. At least one implementation, found on the Qualcomm Technologies QDF2432, only supports 32 bit accesses. SBSAv3, which describes supported access sizes in greater detail, explicitly requires support for both 16 and 32 bit accesses to all registers (and 8 bit accesses to some but not all). Therefore, for broad compatibility, simply use 32 bit accessors for the SBSA UART.

So I gave this a try on a Juno and a Midway. Both have a normal PL011, but I changed the DT to advertise an SBSA UART instead.
This worked fine with the 32bit accessors.
Also according to some research on the hardware size at least the current ARM PL011 implementation are totally fine with 32-bit (as well as 16-bit) accesses.
There is some reluctance about whether this is true for _every_ older PL011 implementation, but they are out of scope here, as we are talking about the SBSA only.
So:

Quoted 6 lines

You can add Juno and Midway to the list of tested systems.

Quoted 1 line

Changes new in v2:

Apologies for omitting the v2 prefix in the second version of the patch that I sent out.

Quoted 17 lines
Quoted 17 lines

* Fixed from address
* Elaborated on forward (SBSAv3) compatibility in commit message
* Included Mark Langsdorf's Tested-by, which now covers:

    QDF2432
    Seattle
    X-Gene 1
---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +0 -0
index c0da0cc..ffb5eb8 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
Quoted 13 lines
Quoted 7 lines

drivers/tty/serial/amba-pl011.c static struct vendor_data vendor_arm = { · line 127

Quoted 7 lines
drivers/tty/serial/amba-pl011.c +1 -0
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,

Do you consider this patch suitable to be included in a 4.6 release candidate? It fixes an issue running this driver on certain hardware, and with gracious assistance we've performed due diligence to check that it does not adversely affect the driver running on other hardware. Would it be useful to send a v3 collecting the acks and tested-bys?

If this isn't a bug fix or regression fix, it's not ok for 4.6-final, it will have to wait for 4.7-rc1.

And yes, collecting the acks and tested-bys would be great, I'm always glad to see that, it makes my job easier. Now that 4.6-rc1 is out, I'll start to dig through the list of pending patches here, give me a few weeks to get all of them, especially due to the conference travel I'm currently doing...

thanks,

greg k-h

#8 lore raw
Quoted 2 lines · Greg Kroah-Hartman

If this isn't a bug fix or regression fix, it's not ok for 4.6-final, it will have to wait for 4.7-rc1.

It fixes a problem on our platform (QDF2432). Without this patch, we can't use the PL011 at all.

Signature · 3 lines
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora
Forum, a Linux Foundation collaborative project.
#9 lore raw
Quoted 6 lines · Timur Tabi
Quoted 2 lines · Greg Kroah-Hartman

If this isn't a bug fix or regression fix, it's not ok for 4.6-final, it will have to wait for 4.7-rc1.

It fixes a problem on our platform (QDF2432). Without this patch, we can't use the PL011 at all.

Did it ever work before? Or is this new functionality?

#10 lore raw
Quoted 9 lines · Greg Kroah-Hartman
Quoted 6 lines · Timur Tabi
Quoted 2 lines · Greg Kroah-Hartman

If this isn't a bug fix or regression fix, it's not ok for 4.6-final, it will have to wait for 4.7-rc1.

It fixes a problem on our platform (QDF2432). Without this patch, we can't use the PL011 at all.

Did it ever work before? Or is this new functionality?

No, it never worked before, so it's not a regression. I guess it all depends on how you define "fix". The driver loads and attempts to use the hardware, but it fails without this patch. The system locks up completely (I guess it throws an unhandled exception or something).

I guess if you take a very limited definition of "fix", then this isn't a fix. I can understand if you didn't want to take it for 4.6-rc7 or something, but for 4.6-rc2, I don't think it's inappropriate.

That's my two cents. We'd like to see it in 4.6-rc2, but the decision is yours.

Signature · 3 lines
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora
Forum, a Linux Foundation collaborative project.
#11 lore raw

PATCH v3 1/1 tty: amba-pl011: Use 32-bit accesses for SBSA UART

In reply to Greg Kroah-Hartman

Version 2 of the Server Base System Architecture (SBSAv2) describes UART hardware registers as 32 bits wide, giving no guidance on access sizes. The SBSA UART driver previously assumed partial-length 16 and 8 bit accesses would work. But the SBSAv2 UART hardware on the Qualcomm Technologies QDF2432 only supports full-length 32 bit register accesses, so use those exclusively. This is compatible with SBSAv3, which explicitly requires UART hardware support 32 (and 16 and sometimes 8) bit accesses.

Tested on Juno, Midway, QDF2432, Seattle, and X-Gene 1.

---
Changes since v2:
* Updated commit message
* Added Andre's Acked-by and Tested-by.

---
 drivers/tty/serial/amba-pl011.c | 1 +
 1 file changed, 1 insertion(+)
drivers/tty/serial/amba-pl011.c +1 -0
index 7c198e0..a2aa655 100644
@@ -121,6 +121,7 @@ static struct vendor_data vendor_arm = {
 
 static struct vendor_data vendor_sbsa = {
 	.reg_offset		= pl011_std_offsets,
+	.access_32b		= true,
 	.oversampling		= false,
 	.dma_threshold		= false,
 	.cts_event_workaround	= false,
Signature · 3 lines
Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project