Re: Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
Hi, I guess the kernel will be rebuilt due to the ARM64 breakage but I confirm it works fine on the following machines : Intel Core i7 4790 w/ GeForce GTX 1070 and NVidia BLOB 384.111 Intel Core i7 4800 MQ Intel Pentium N3700 AMD Phenom x4 9840 w/ GeForce GTX 970 and NVidia BLOB 384.111 However it seems that the updated kernel does not fix against variant 1 and 2 of Spectre. :'( Rgds
Re: Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Fri, 2018-02-09 at 14:00 +0100, Julien Aubin wrote: > I guess the kernel will be rebuilt due to the ARM64 breakage Yes, I'll push a -2 with ABI change ignored on arm64. > > but I confirm it works fine on the following machines : > Intel Core i7 4790 w/ GeForce GTX 1070 and NVidia BLOB 384.111 > Intel Core i7 4800 MQ > Intel Pentium N3700 > AMD Phenom x4 9840 w/ GeForce GTX 970 and NVidia BLOB 384.111 Thanks for the report. > > However it seems that the updated kernel does not fix against variant 1 and > 2 of Spectre. :'( Yes indeed. Retpoline is included in the kernel but there is not gcc support yet. And IBRS/IBPB is not there yet, nor the microcodes updates. Regards, -- Yves-Alexis signature.asc Description: This is a digitally signed message part
Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Sun, 2018-01-28 at 20:34 +0100, Yves-Alexis Perez wrote: > > No, that's not necessary. But could you wait a day or two so I have a > > chance to review your work? > > Sure, I'd welcome a review on this actually :) Hi Ben, 4.9.80-1 is ready for me. I'm not totally sure how to interpret: 2018/02/02 - [12:40:22] (bwh): Corsac: Sorry, no, why don't you go ahead anyway Is it go for upload, even without review? If so, I'll do the 4.9.80 upload tonight and watch the list for any bug from stretch-pu. Regards, -- Yves-Alexis signature.asc Description: This is a digitally signed message part
Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Sun, 2018-01-28 at 18:45 +, Ben Hutchings wrote: > > So you want to include the patch even if it's reverted upstream? > > Oh, I misunderstood what you were saying. No I don't. But I think it > makes sense to trigger a rebuild of modules once the retpoline compiler > support is in place. Agreed. > Do > > I need to ping someone before doing the upload? > > No, that's not necessary. But could you wait a day or two so I have a > chance to review your work? Sure, I'd welcome a review on this actually :) Regards, -- Yves-Alexis signature.asc Description: This is a digitally signed message part
Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Sun, 2018-01-28 at 19:27 +0100, Yves-Alexis Perez wrote: > On Sun, 2018-01-28 at 18:24 +, Ben Hutchings wrote: > > > So I went ahead an updated to 4.9.78 and added the revert for the VERMAGIC > > > stuff (which is queued for 4.9.79). > > > > This makes sense. I think once an updated gcc is available in stretch > > we should bump the ABI number and Build-Depends at the same time as un- > > reverting that. > > So you want to include the patch even if it's reverted upstream? Oh, I misunderstood what you were saying. No I don't. But I think it makes sense to trigger a rebuild of modules once the retpoline compiler support is in place. > > > I'm currently running a full build in pbuilder in a stretch chroot, so it > > > should be ready for upload if it's ok for you. > > > > > > I'm unsure about stable-uploads procedure for the kernel (SRM and debian- > > > > boot > > > notifications for example) so pointer to any specific stuff would be > > > > welcome. > > > > I've tended to upload shortly before point releases, but I think we > > should really be uploading earlier and more often so regressions are > > more lilely to be caught before a point release. > > Yes I think it'd be a good idea to let this one stay a bit longer in s-p-u. Do > I need to ping someone before doing the upload? No, that's not necessary. But could you wait a day or two so I have a chance to review your work? Ben. -- Ben Hutchings If the facts do not conform to your theory, they must be disposed of. signature.asc Description: This is a digitally signed message part
Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Sun, 2018-01-28 at 18:24 +, Ben Hutchings wrote: > > So I went ahead an updated to 4.9.78 and added the revert for the VERMAGIC > > stuff (which is queued for 4.9.79). > > This makes sense. I think once an updated gcc is available in stretch > we should bump the ABI number and Build-Depends at the same time as un- > reverting that. So you want to include the patch even if it's reverted upstream? > > > I'm currently running a full build in pbuilder in a stretch chroot, so it > > should be ready for upload if it's ok for you. > > > > I'm unsure about stable-uploads procedure for the kernel (SRM and debian- > boot > > notifications for example) so pointer to any specific stuff would be > welcome. > > I've tended to upload shortly before point releases, but I think we > should really be uploading earlier and more often so regressions are > more lilely to be caught before a point release. Yes I think it'd be a good idea to let this one stay a bit longer in s-p-u. Do I need to ping someone before doing the upload? Regards, -- Yves-Alexis signature.asc Description: This is a digitally signed message part
Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Sun, 2018-01-28 at 18:42 +0100, Yves-Alexis Perez wrote: > On Wed, 2018-01-24 at 13:52 +0100, Yves-Alexis Perez wrote: > > > > work on 4.9.77 is mostly done, so yes I'd like to push it to stretch before > > next point relase. 4.9.78 is just out but I'm unsure if we want to hold it > > or > > not. > > Hi Ben, > > So I went ahead an updated to 4.9.78 and added the revert for the VERMAGIC > stuff (which is queued for 4.9.79). This makes sense. I think once an updated gcc is available in stretch we should bump the ABI number and Build-Depends at the same time as un- reverting that. > I'm currently running a full build in pbuilder in a stretch chroot, so it > should be ready for upload if it's ok for you. > > I'm unsure about stable-uploads procedure for the kernel (SRM and debian-boot > notifications for example) so pointer to any specific stuff would be welcome. I've tended to upload shortly before point releases, but I think we should really be uploading earlier and more often so regressions are more lilely to be caught before a point release. Ben. -- Ben Hutchings If the facts do not conform to your theory, they must be disposed of. signature.asc Description: This is a digitally signed message part
Re: Bug#888263: Spectre : release kernel 4.9.77 to stretch before p-u
On Wed, 2018-01-24 at 13:52 +0100, Yves-Alexis Perez wrote: > > work on 4.9.77 is mostly done, so yes I'd like to push it to stretch before > next point relase. 4.9.78 is just out but I'm unsure if we want to hold it or > not. Hi Ben, So I went ahead an updated to 4.9.78 and added the revert for the VERMAGIC stuff (which is queued for 4.9.79). I'm currently running a full build in pbuilder in a stretch chroot, so it should be ready for upload if it's ok for you. I'm unsure about stable-uploads procedure for the kernel (SRM and debian-boot notifications for example) so pointer to any specific stuff would be welcome. Regards, -- Yves-Alexis signature.asc Description: This is a digitally signed message part