Hi Russell,
On 27/10/2023 14:46, Russell King (Oracle) wrote:
On Fri, Oct 20, 2023 at 04:56:20PM +0000, Salil Mehta wrote:
From: Russell King linux@armlinux.org.uk Sent: Friday, October 20, 2023 5:48 PM To: Salil Mehta salil.mehta@huawei.com Cc: James Morse james.morse@arm.com; linaro-open-discussions@op- lists.linaro.org; Joyce Qi joyce.qi@linaro.org; Lorenzo Pieralisi lorenzo.pieralisi@linaro.org; Jonathan Cameron jonathan.cameron@huawei.com; Shameerali Kolothum Thodi shameerali.kolothum.thodi@huawei.com; karl.heubaum@oracle.com; darren@os.amperecomputing.com; ilkka@os.amperecomputing.com; miguel.luis@oracle.com; vishnu@os.amperecomputing.com; Linuxarm linuxarm@huawei.com; salil.mehta@opnsrc.net Subject: Re: [Request] Regarding non-RFC patch-set of Virtual CPU Hotplug kernel support
On Fri, Oct 20, 2023 at 04:40:59PM +0000, Salil Mehta wrote:
I'm afraid it isn't going to happen - going through the review comments there's some ambiguities there (at least to me) that I can't solve without input from the reviewers. I haven't even managed to get half way through the patches yet.
Possible to share the link of the specific review comments and the patches you are referring to here?
I'm preferring to the comments on the RFC v2 posting on the 13th September by James. I think you already provided a link.
https://lore.kernel.org/linux-arm-kernel/20230913163823.7880-1-james.morse@a...
IICRC, you mentioned that this patch-set was ready for submission?
In case you missed it, Rafael Wysocki isn't happy with various patches in this series - we have his comments on one of the patches which sound like the approach with the "enabled" bit won't work - at the very least in an arch-independent context. He also says he has other concerns, but hasn't set out what those are, and says he will do so after the merge window.
He summarises it as "and it doesn't look good" which rings alarm bells. I am hoping this doesn't mean it's back to the drawing board for this feature.
The angle of attack on this is the enabled bit has always existed, we didn't invent anything new here. Linux has historically ignored it. Its fine to say "its always junk on x86" if we don't trust vendors to get this right.
I'd much prefer we quirk the devices where this is discovered to be junk, instead of declaring outright that we ignore this bit of the spec.
I'm also wondering why it's taken Rafael this long to state this concern - it is not like this has changed between RFC v2 and the series I posted, and Rafael did comment on the RFC v2 series. So I don't really understand why it's taken so long to bring up this concern.
Hopefully we will find out more after the merge window is over in about two weeks time.
Salil - I suspect this means that the qemu patches need to be held as well, because if the kernel side approach has to change it may impact the qemu side as well.