Message ID | 20230925183922.71728-1-john.cabaj@canonical.com |
---|---|
Headers | show |
Series | SEV_SNP failure to init | expand |
On 9/25/23 12:39 PM, John Cabaj wrote: > BugLink: https://bugs.launchpad.net/bugs/2037316 > > [Impact] > > * Kernel fails to boot on SEV-SNP instances when compiled with GCC 12.3.0 > > [Fix] > > * https://lore.kernel.org/lkml/20230912002703.3924521-1-acdunlap@google.com/ > > [Test Case] > > * Compile tested > * Boot tested > * Tested by Google > > [Where things could go wrong] > > * Patches relatively isolated and maintain similar checking functionality, > just earlier in boot. Likely a low chance of regression. > > Adam Dunlap (2): > x86/sev-es: Allow copy_from_kernel_nofault in earlier boot > x86/sev-es: Only set x86_virt_bits to correct value > > arch/x86/kernel/cpu/common.c | 37 +++++++++++++++++++++--------------- > arch/x86/mm/maccess.c | 19 +++++++++++++----- > 2 files changed, 36 insertions(+), 20 deletions(-) > Is there any reason these patches shouldn't be applied to the generic kernel ? That way all the derivatives are fixed wrt SEV-SNP.
On 9/25/23 1:47 PM, Tim Gardner wrote: > On 9/25/23 12:39 PM, John Cabaj wrote: >> BugLink: https://bugs.launchpad.net/bugs/2037316 >> >> [Impact] >> >> * Kernel fails to boot on SEV-SNP instances when compiled with GCC 12.3.0 >> >> [Fix] >> >> * https://lore.kernel.org/lkml/20230912002703.3924521-1-acdunlap@google.com/ >> >> [Test Case] >> >> * Compile tested >> * Boot tested >> * Tested by Google >> >> [Where things could go wrong] >> >> * Patches relatively isolated and maintain similar checking functionality, >> just earlier in boot. Likely a low chance of regression. >> >> Adam Dunlap (2): >> x86/sev-es: Allow copy_from_kernel_nofault in earlier boot >> x86/sev-es: Only set x86_virt_bits to correct value >> >> arch/x86/kernel/cpu/common.c | 37 +++++++++++++++++++++--------------- >> arch/x86/mm/maccess.c | 19 +++++++++++++----- >> 2 files changed, 36 insertions(+), 20 deletions(-) >> > > Is there any reason these patches shouldn't be applied to the generic kernel ? That way all the derivatives are fixed wrt SEV-SNP. They're simple enough that we could apply to the generic kernel as well, since I know a couple other kernels could use it. GCP needs this for 2023.10.02, so these patches would need ACKs and application by 27-Sep. Might need some coordination with the Stable team. John
On 9/25/23 12:51 PM, John Cabaj wrote: > On 9/25/23 1:47 PM, Tim Gardner wrote: >> On 9/25/23 12:39 PM, John Cabaj wrote: >>> BugLink: https://bugs.launchpad.net/bugs/2037316 >>> >>> [Impact] >>> >>> * Kernel fails to boot on SEV-SNP instances when compiled with GCC 12.3.0 >>> >>> [Fix] >>> >>> * https://lore.kernel.org/lkml/20230912002703.3924521-1-acdunlap@google.com/ >>> >>> [Test Case] >>> >>> * Compile tested >>> * Boot tested >>> * Tested by Google >>> >>> [Where things could go wrong] >>> >>> * Patches relatively isolated and maintain similar checking functionality, >>> just earlier in boot. Likely a low chance of regression. >>> >>> Adam Dunlap (2): >>> x86/sev-es: Allow copy_from_kernel_nofault in earlier boot >>> x86/sev-es: Only set x86_virt_bits to correct value >>> >>> arch/x86/kernel/cpu/common.c | 37 +++++++++++++++++++++--------------- >>> arch/x86/mm/maccess.c | 19 +++++++++++++----- >>> 2 files changed, 36 insertions(+), 20 deletions(-) >>> >> >> Is there any reason these patches shouldn't be applied to the generic kernel ? That way all the derivatives are fixed wrt SEV-SNP. > > They're simple enough that we could apply to the generic kernel as well, since I know a couple other kernels could use it. GCP needs this for 2023.10.02, so these patches would need ACKs and application by 27-Sep. Might need some coordination with the Stable team. > > > John > Acked-by: Tim Gardner <tim.gardner@canonical.com> Would you mind reposting these for generic ? I'd like to get them in the 2023.10.02 cycle as well. Should these be SAUCE until they've made it upstream ?
On 9/25/23 2:19 PM, Tim Gardner wrote: > On 9/25/23 12:51 PM, John Cabaj wrote: >> On 9/25/23 1:47 PM, Tim Gardner wrote: >>> On 9/25/23 12:39 PM, John Cabaj wrote: >>>> BugLink: https://bugs.launchpad.net/bugs/2037316 >>>> >>>> [Impact] >>>> >>>> * Kernel fails to boot on SEV-SNP instances when compiled with GCC 12.3.0 >>>> >>>> [Fix] >>>> >>>> * https://lore.kernel.org/lkml/20230912002703.3924521-1-acdunlap@google.com/ >>>> >>>> [Test Case] >>>> >>>> * Compile tested >>>> * Boot tested >>>> * Tested by Google >>>> >>>> [Where things could go wrong] >>>> >>>> * Patches relatively isolated and maintain similar checking functionality, >>>> just earlier in boot. Likely a low chance of regression. >>>> >>>> Adam Dunlap (2): >>>> x86/sev-es: Allow copy_from_kernel_nofault in earlier boot >>>> x86/sev-es: Only set x86_virt_bits to correct value >>>> >>>> arch/x86/kernel/cpu/common.c | 37 +++++++++++++++++++++--------------- >>>> arch/x86/mm/maccess.c | 19 +++++++++++++----- >>>> 2 files changed, 36 insertions(+), 20 deletions(-) >>>> >>> >>> Is there any reason these patches shouldn't be applied to the generic kernel ? That way all the derivatives are fixed wrt SEV-SNP. >> >> They're simple enough that we could apply to the generic kernel as well, since I know a couple other kernels could use it. GCP needs this for 2023.10.02, so these patches would need ACKs and application by 27-Sep. Might need some coordination with the Stable team. >> >> >> John >> > Acked-by: Tim Gardner <tim.gardner@canonical.com> > > Would you mind reposting these for generic ? I'd like to get them in the 2023.10.02 cycle as well. Will do, with below considerations. > > Should these be SAUCE until they've made it upstream ? Will fix this in the APPLIED. Thanks, John
On Mon, Sep 25, 2023 at 01:19:51PM -0600, Tim Gardner wrote: > On 9/25/23 12:51 PM, John Cabaj wrote: > > On 9/25/23 1:47 PM, Tim Gardner wrote: > > > On 9/25/23 12:39 PM, John Cabaj wrote: > > > > BugLink: https://bugs.launchpad.net/bugs/2037316 > > > > > > > > [Impact] > > > > > > > > * Kernel fails to boot on SEV-SNP instances when compiled with GCC 12.3.0 > > > > > > > > [Fix] > > > > > > > > * https://lore.kernel.org/lkml/20230912002703.3924521-1-acdunlap@google.com/ > > > > > > > > [Test Case] > > > > > > > > * Compile tested > > > > * Boot tested > > > > * Tested by Google > > > > > > > > [Where things could go wrong] > > > > > > > > * Patches relatively isolated and maintain similar checking functionality, > > > > just earlier in boot. Likely a low chance of regression. > > > > > > > > Adam Dunlap (2): > > > > x86/sev-es: Allow copy_from_kernel_nofault in earlier boot > > > > x86/sev-es: Only set x86_virt_bits to correct value > > > > > > > > arch/x86/kernel/cpu/common.c | 37 +++++++++++++++++++++--------------- > > > > arch/x86/mm/maccess.c | 19 +++++++++++++----- > > > > 2 files changed, 36 insertions(+), 20 deletions(-) > > > > > > > > > > Is there any reason these patches shouldn't be applied to the generic kernel ? That way all the derivatives are fixed wrt SEV-SNP. > > > > They're simple enough that we could apply to the generic kernel as well, since I know a couple other kernels could use it. GCP needs this for 2023.10.02, so these patches would need ACKs and application by 27-Sep. Might need some coordination with the Stable team. > > > > > > John > > > Acked-by: Tim Gardner <tim.gardner@canonical.com> > > Would you mind reposting these for generic ? I'd like to get them in the > 2023.10.02 cycle as well. Those are not upstream yet, so I'd be weary to apply them so soon in -generic. If other kernels seem to require it, perhaps you should consider sending the patches for those kernels instead of having them land on -generic so soon. Cascardo. > > Should these be SAUCE until they've made it upstream ? > -- > ----------- > Tim Gardner > Canonical, Inc
Acked-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Don't forget to apply them as SAUCE.
Cascardo.
On 9/25/23 1:39 PM, John Cabaj wrote: > BugLink: https://bugs.launchpad.net/bugs/2037316 > > [Impact] > > * Kernel fails to boot on SEV-SNP instances when compiled with GCC 12.3.0 > > [Fix] > > * https://lore.kernel.org/lkml/20230912002703.3924521-1-acdunlap@google.com/ > > [Test Case] > > * Compile tested > * Boot tested > * Tested by Google > > [Where things could go wrong] > > * Patches relatively isolated and maintain similar checking functionality, > just earlier in boot. Likely a low chance of regression. > > Adam Dunlap (2): > x86/sev-es: Allow copy_from_kernel_nofault in earlier boot > x86/sev-es: Only set x86_virt_bits to correct value > > arch/x86/kernel/cpu/common.c | 37 +++++++++++++++++++++--------------- > arch/x86/mm/maccess.c | 19 +++++++++++++----- > 2 files changed, 36 insertions(+), 20 deletions(-) > Patches applied to mantic:linux-gcp and lunar:linux-gcp master-next branches. Thanks, John