Skip to contentSkip to conversation

Unreviewed arm64 ABI change in linux-next via driver-core tree

DISCUSSION Will Deacon 4 participants 5 messages Sep 26, 2017

lore
Share & tools
lore
#0 lore raw

Hi Greg,

We stumbled over a change in linux-next that has ABI implications for arm64 and, as far as we can tell, has not been reviewed:

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/arm64/kernel?id=fb7ff3f8721b87bf078868c9a252fa0cf71a47c2

("fb7ff3f8721b cpuinfo: implement sysfs nodes for arm64")

It looks like this was requested by you on LKML here:

https://lkml.org/lkml/2017/8/2/663

but these sorts of changes really need to go on the linux-arm-kernel mailing list with the maintainers on cc so that the patches can be properly reviewed and eventually acked if appropriate. In this case, I can't find any trace of the patch on any of the mailing lists to which I'm subscribed.

Please can you drop this from your tree until we've had a chance to review it properly? If Felix can repost it as above, then we can get the discussion started because, aside from this email, we have concerns about exactly what is being exposed to userspace here.

Thanks,

Will

#1 lore raw

Hi Will, Hi Greg,

It seems this is all my fault. It looks like forgot to check that the outgoing mailserver send the patches correctly to the mailinglists which doesn't happen.

So to make this horrible situation short maybe drop the patches and I will resend them and make sure this time everything works as expected

Sorry Greg for all the stress :(

Felix

Quoted 18 lines · Will Deacon

Hi Greg,

We stumbled over a change in linux-next that has ABI implications for arm64 and, as far as we can tell, has not been reviewed:

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/arm64/kernel?id=fb7ff3f8721b87bf078868c9a252fa0cf71a47c2

("fb7ff3f8721b cpuinfo: implement sysfs nodes for arm64")

It looks like this was requested by you on LKML here:

https://lkml.org/lkml/2017/8/2/663

but these sorts of changes really need to go on the linux-arm-kernel mailing list with the maintainers on cc so that the patches can be properly reviewed and eventually acked if appropriate. In this case, I can't find any trace of the patch on any of the mailing lists to which I'm subscribed.

Quoted 9 lines

Please can you drop this from your tree until we've had a chance to review it properly? If Felix can repost it as above, then we can get the discussion started because, aside from this email, we have concerns about exactly what is being exposed to userspace here.

Thanks,

Will

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20170926/cbb9c5f3/attachment.sig>

#2 lore raw
In reply to Will Deacon

Hi,

Quoted 5 lines · Will Deacon

Hi Greg,

We stumbled over a change in linux-next that has ABI implications for arm64 and, as far as we can tell, has not been reviewed:

Quoted 8 lines

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/arm64/kernel?id=fb7ff3f8721b87bf078868c9a252fa0cf71a47c2

("fb7ff3f8721b cpuinfo: implement sysfs nodes for arm64")

Please can you drop this from your tree until we've had a chance to review it properly? If Felix can repost it as above, then we can get the discussion started because, aside from this email, we have concerns about exactly what is being exposed to userspace here.

Seconded.

Experience with the existing /proc/cpuinfo shows that this needs *very* careful review.

I've worked on arm64's /proc/cpuinfo, and I'd appreciate being Cc'd on future postings of this series. I'd also recommend that future postings are also Cc'd to linux-arch (and maybe linux-api) so that the interface is reviewed by a larger set of relevant parties.

Thanks,
Mark.

#3 lore raw
Quoted 19 lines · Mark Rutland

Hi,

Quoted 5 lines · Will Deacon

Hi Greg,

We stumbled over a change in linux-next that has ABI implications for arm64 and, as far as we can tell, has not been reviewed:

Quoted 8 lines

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/arm64/kernel?id=fb7ff3f8721b87bf078868c9a252fa0cf71a47c2

("fb7ff3f8721b cpuinfo: implement sysfs nodes for arm64")

Please can you drop this from your tree until we've had a chance to review it properly? If Felix can repost it as above, then we can get the discussion started because, aside from this email, we have concerns about exactly what is being exposed to userspace here.

Seconded.

Quoted 7 lines

Experience with the existing /proc/cpuinfo shows that this needs *very* careful review.

I've worked on arm64's /proc/cpuinfo, and I'd appreciate being Cc'd on future postings of this series. I'd also recommend that future postings are also Cc'd to linux-arch (and maybe linux-api) so that the interface is reviewed by a larger set of relevant parties.

I will do that!

Quoted 3 lines

Thanks,
Mark.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20170926/eaad4b39/attachment.sig>

#4 lore raw
In reply to Mark Rutland
Quoted 10 lines · Mark Rutland

Hi,

Quoted 7 lines · Will Deacon

Hi Greg,

We stumbled over a change in linux-next that has ABI implications for arm64 and, as far as we can tell, has not been reviewed:

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/arm64/kernel?id=fb7ff3f8721b87bf078868c9a252fa0cf71a47c2

Quoted 8 lines
Quoted 6 lines

("fb7ff3f8721b cpuinfo: implement sysfs nodes for arm64")

Please can you drop this from your tree until we've had a chance to review it properly? If Felix can repost it as above, then we can get the discussion started because, aside from this email, we have concerns about exactly what is being exposed to userspace here.

Seconded.

Now dropped.

Quoted 2 lines

Experience with the existing /proc/cpuinfo shows that this needs *very* careful review.

Hah, yes, but with a sysfs interface, one would "hope" that you can get it right this time around :)

Quoted 4 lines

I've worked on arm64's /proc/cpuinfo, and I'd appreciate being Cc'd on future postings of this series. I'd also recommend that future postings are also Cc'd to linux-arch (and maybe linux-api) so that the interface is reviewed by a larger set of relevant parties.

For some reason, this set of patches never made it to lkml or the lists multiple times. Both you all, and the x86 developers have objected to these changes, so they are gone for now.

Felix, please resend the series and cc: the needed people.

thanks,

greg k-h