Skip to contentSkip to conversation

U-Boot v2022.01-rc3 released

ANN Tom Rini 2 participants 6 messages Nov 29, 2021 – Jul 6, 2022

lore
Share & tools
lore

Tom Rini

author
#0 lore raw

Hey all,

It's been two weeks since v2022.01-rc2, so here's -rc3.

To repeat what I said with -rc2, we've enabled issue tracking on our gitlab instance. You can sign up and then be able to file issues at:
https://source.denx.de/groups/u-boot/-/issues

This is intended for everyone to be able to use, both custodians for their own needs (you can see for example Heinrich has filed something for UEFI and LMB) as well as users to just report bugs so they don't feel like they're lost in the mailing list.

As noted with the last release, the -next branch is open and I'll sync in -rc3 shortly. Please feel free to get a PR ready now if you're able.

In terms of a changelog,
git log --merges v2022.01-rc2..v2022.01-rc3
contains what I've pulled but as always, better PR messages and tags will provide better results here.

So we're now looking at regular releases every other Monday, and with final release on January 10th, 2022. Thanks all!

Signature · 1 line
Tom
#1 lore raw
Quoted 16 lines · Tom Rini

Hey all,

It's been two weeks since v2022.01-rc2, so here's -rc3.

To repeat what I said with -rc2, we've enabled issue tracking on our gitlab instance. You can sign up and then be able to file issues at:
https://source.denx.de/groups/u-boot/-/issues

This is intended for everyone to be able to use, both custodians for their own needs (you can see for example Heinrich has filed something for UEFI and LMB) as well as users to just report bugs so they don't feel like they're lost in the mailing list.

As noted with the last release, the -next branch is open and I'll sync in -rc3 shortly. Please feel free to get a PR ready now if you're able.

Quoted 8 lines

In terms of a changelog,
git log --merges v2022.01-rc2..v2022.01-rc3
contains what I've pulled but as always, better PR messages and tags will provide better results here.

So we're now looking at regular releases every other Monday, and with final release on January 10th, 2022. Thanks all!

Hey tom,

Thank you for the update as always!

I have a question about a recent commit, I hope its okay to ask here.
In commit cd82f199852d88218e1f17f5ec07cdd9112a89c4
In arch/arm/lib/relocate.S:81 on my SBC it returns an invalid value.
My soc is Thumb2 but the instruction adr r3, relocate_code
assembles to subw r3, pc, #3 which is not 32bit aligned. If i change the instruction to adr.w r3, relocate_code it evaluates to subw r3, pc, #4, which is.

There is a slight problem as it seems to work fine on my laptop using Debian bullseye, but on my Desktop where I found this I'm running sid. They are both gcc-10. I have yet to find a way to consistently replicate it.

What are your thoughts of this?

Thanks,
Jesse

author
#2 lore raw
Quoted 39 lines · Jesse Taube
Quoted 16 lines · Tom Rini

Hey all,

It's been two weeks since v2022.01-rc2, so here's -rc3.

To repeat what I said with -rc2, we've enabled issue tracking on our gitlab instance. You can sign up and then be able to file issues at:
https://source.denx.de/groups/u-boot/-/issues

This is intended for everyone to be able to use, both custodians for their own needs (you can see for example Heinrich has filed something for UEFI and LMB) as well as users to just report bugs so they don't feel like they're lost in the mailing list.

As noted with the last release, the -next branch is open and I'll sync in -rc3 shortly. Please feel free to get a PR ready now if you're able.

Quoted 8 lines

In terms of a changelog,
git log --merges v2022.01-rc2..v2022.01-rc3
contains what I've pulled but as always, better PR messages and tags will provide better results here.

So we're now looking at regular releases every other Monday, and with final release on January 10th, 2022. Thanks all!

Hey tom,

Thank you for the update as always!

I have a question about a recent commit, I hope its okay to ask here. In commit cd82f199852d88218e1f17f5ec07cdd9112a89c4
In arch/arm/lib/relocate.S:81 on my SBC it returns an invalid value. My soc is Thumb2 but the instruction adr r3, relocate_code assembles to subw r3, pc, #3 which is not 32bit aligned. If i change the instruction to adr.w r3, relocate_code it evaluates to subw r3, pc, #4, which is.

Quoted 5 lines

There is a slight problem as it seems to work fine on my laptop using Debian bullseye, but on my Desktop where I found this I'm running sid. They are both gcc-10. I have yet to find a way to consistently replicate it.

What are your thoughts of this?

It sounds like you have enough details here to file a bug with Debian and see where that leads.

Signature · 1 line
Tom
#3 lore raw
Quoted 40 lines · Tom Rini
Quoted 39 lines · Jesse Taube
Quoted 16 lines · Tom Rini

Hey all,

It's been two weeks since v2022.01-rc2, so here's -rc3.

To repeat what I said with -rc2, we've enabled issue tracking on our gitlab instance. You can sign up and then be able to file issues at:
https://source.denx.de/groups/u-boot/-/issues

This is intended for everyone to be able to use, both custodians for their own needs (you can see for example Heinrich has filed something for UEFI and LMB) as well as users to just report bugs so they don't feel like they're lost in the mailing list.

As noted with the last release, the -next branch is open and I'll sync in -rc3 shortly. Please feel free to get a PR ready now if you're able.

Quoted 8 lines

In terms of a changelog,
git log --merges v2022.01-rc2..v2022.01-rc3
contains what I've pulled but as always, better PR messages and tags will provide better results here.

So we're now looking at regular releases every other Monday, and with final release on January 10th, 2022. Thanks all!

Hey tom,

Thank you for the update as always!

I have a question about a recent commit, I hope its okay to ask here. In commit cd82f199852d88218e1f17f5ec07cdd9112a89c4
In arch/arm/lib/relocate.S:81 on my SBC it returns an invalid value. My soc is Thumb2 but the instruction adr r3, relocate_code assembles to subw r3, pc, #3 which is not 32bit aligned. If i change the instruction to adr.w r3, relocate_code it evaluates to subw r3, pc, #4, which is.

Quoted 9 lines
Quoted 5 lines

There is a slight problem as it seems to work fine on my laptop using Debian bullseye, but on my Desktop where I found this I'm running sid. They are both gcc-10. I have yet to find a way to consistently replicate it.

What are your thoughts of this?

It sounds like you have enough details here to file a bug with Debian and see where that leads.

Sorry for replying to old messages but the bug has appeared again in a separate case the bug seems to be introduced in
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=d3e52e120b68bf19552743fbc078e0a759f48cb7

The test code is as follows

```
.syntax unified 

.global bug; 

.align 4 

bug: 

        adr     r3, bug 

.size bug, .-bug 

.type bug 2; // This changes offset from 4 to 3 in include/linux/linkage.h:ENDPROC
//arm-linux-gnueabi-as  -march=armv7-m -c -o bug.o bug.S && arm-linux-gnueabi-objdump --disassemble=bug bug.o
```

Releases >2.37 have this issue and causes some arm platforms in u-boot at /arch/arm/lib/relocate.S:81 adr r3, relocate_code

I'm wondering if there should be a mitigation for this till binutils is patched.

Thanks,
Jesse

author
#4 lore raw
Quoted 75 lines · Jesse Taube
Quoted 40 lines · Tom Rini
Quoted 39 lines · Jesse Taube
Quoted 16 lines · Tom Rini

Hey all,

It's been two weeks since v2022.01-rc2, so here's -rc3.

To repeat what I said with -rc2, we've enabled issue tracking on our gitlab instance. You can sign up and then be able to file issues at:
https://source.denx.de/groups/u-boot/-/issues

This is intended for everyone to be able to use, both custodians for their own needs (you can see for example Heinrich has filed something for UEFI and LMB) as well as users to just report bugs so they don't feel like they're lost in the mailing list.

As noted with the last release, the -next branch is open and I'll sync in -rc3 shortly. Please feel free to get a PR ready now if you're able.

Quoted 8 lines

In terms of a changelog,
git log --merges v2022.01-rc2..v2022.01-rc3
contains what I've pulled but as always, better PR messages and tags will provide better results here.

So we're now looking at regular releases every other Monday, and with final release on January 10th, 2022. Thanks all!

Hey tom,

Thank you for the update as always!

I have a question about a recent commit, I hope its okay to ask here. In commit cd82f199852d88218e1f17f5ec07cdd9112a89c4
In arch/arm/lib/relocate.S:81 on my SBC it returns an invalid value. My soc is Thumb2 but the instruction adr r3, relocate_code assembles to subw r3, pc, #3 which is not 32bit aligned. If i change the instruction to adr.w r3, relocate_code it evaluates to subw r3, pc, #4, which is.

Quoted 9 lines
Quoted 5 lines

There is a slight problem as it seems to work fine on my laptop using Debian bullseye, but on my Desktop where I found this I'm running sid. They are both gcc-10. I have yet to find a way to consistently replicate it.

What are your thoughts of this?

It sounds like you have enough details here to file a bug with Debian and see where that leads.

Sorry for replying to old messages but the bug has appeared again in a separate case the bug seems to be introduced in
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=d3e52e120b68bf19552743fbc078e0a759f48cb7

The test code is as follows

```
.syntax unified

.global bug;

.align 4

bug:

        adr     r3, bug

.size bug, .-bug

.type bug 2; // This changes offset from 4 to 3 in
include/linux/linkage.h:ENDPROC
//arm-linux-gnueabi-as  -march=armv7-m -c -o bug.o bug.S &&
arm-linux-gnueabi-objdump --disassemble=bug bug.o
```
Quoted 5 lines

Releases >2.37 have this issue and causes some arm platforms in u-boot at
/arch/arm/lib/relocate.S:81 adr r3, relocate_code

I'm wondering if there should be a mitigation for this till binutils is patched.

Thanks for being persistent on this. What would a mitigation look like?

Signature · 1 line
Tom
#5 lore raw
Quoted 79 lines · Tom Rini
Quoted 78 lines · Jesse Taube
Quoted 40 lines · Tom Rini
Quoted 39 lines · Jesse Taube
Quoted 16 lines · Tom Rini

Hey all,

It's been two weeks since v2022.01-rc2, so here's -rc3.

To repeat what I said with -rc2, we've enabled issue tracking on our gitlab instance. You can sign up and then be able to file issues at:
https://source.denx.de/groups/u-boot/-/issues

This is intended for everyone to be able to use, both custodians for their own needs (you can see for example Heinrich has filed something for UEFI and LMB) as well as users to just report bugs so they don't feel like they're lost in the mailing list.

As noted with the last release, the -next branch is open and I'll sync in -rc3 shortly. Please feel free to get a PR ready now if you're able.

Quoted 8 lines

In terms of a changelog,
git log --merges v2022.01-rc2..v2022.01-rc3
contains what I've pulled but as always, better PR messages and tags will provide better results here.

So we're now looking at regular releases every other Monday, and with final release on January 10th, 2022. Thanks all!

Hey tom,

Thank you for the update as always!

I have a question about a recent commit, I hope its okay to ask here. In commit cd82f199852d88218e1f17f5ec07cdd9112a89c4
In arch/arm/lib/relocate.S:81 on my SBC it returns an invalid value. My soc is Thumb2 but the instruction adr r3, relocate_code assembles to subw r3, pc, #3 which is not 32bit aligned. If i change the instruction to adr.w r3, relocate_code it evaluates to subw r3, pc, #4, which is.

Quoted 9 lines
Quoted 5 lines

There is a slight problem as it seems to work fine on my laptop using Debian bullseye, but on my Desktop where I found this I'm running sid. They are both gcc-10. I have yet to find a way to consistently replicate it.

What are your thoughts of this?

It sounds like you have enough details here to file a bug with Debian and see where that leads.

Sorry for replying to old messages but the bug has appeared again in a separate case the bug seems to be introduced in
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=d3e52e120b68bf19552743fbc078e0a759f48cb7

The test code is as follows

```
.syntax unified

.global bug;

.align 4

bug:

         adr     r3, bug

.size bug, .-bug

.type bug 2; // This changes offset from 4 to 3 in
include/linux/linkage.h:ENDPROC
//arm-linux-gnueabi-as  -march=armv7-m -c -o bug.o bug.S &&
arm-linux-gnueabi-objdump --disassemble=bug bug.o
```

Releases >2.37 have this issue and causes some arm platforms in u-boot at
/arch/arm/lib/relocate.S:81 adr r3, relocate_code

Quoted 5 lines
Quoted 2 lines

I'm wondering if there should be a mitigation for this till binutils is patched.

Thanks for being persistent on this. What would a mitigation look like?

Other people have suggested to use subw r3, pc, #4 and ifdef for different variants. We could also do adr.w, what do you recommend. Unfortunately I'm not very familiar with arm asm so I was just putting it out there because many here know more. I'll continue to look into fixes for u-boot. And look into fixes for binutils as 2.37 is now in buildroot and other tool and distros. I'm not sure if line 81 is the only one that is affected but if it is then its a much easier fix than every adr instruction.

Thanks,
Jesse