Re: [PATCH 0/2] i386/a-b-bootblock: zero the first byte of each page on start
27.09.2023, 00:02, "Peter Xu" :On Thu, Sep 07, 2023 at 10:29:42PM +0300, Daniil Tatianin wrote: This series fixes an issue where the outcome of the migration qtest relies on the initial memory contents all being the same across the first 100MiB of RAM, which is a very fragile invariant. We fix this by making sure we zero the first byte of every testable page in range beforehand.What's the difference between this one and:https://lore.kernel.org/r/20230907193051.1609310-1-d-tatia...@yandex-team.ru?There's a v2 version of this series that adds a fix for a similar issue in s390x as well, please look at that one instead.Thanks, --Peter Xu
Re: [PATCH 0/2] i386/a-b-bootblock: zero the first byte of each page on start
27.09.2023, 00:02, "Peter Xu" :On Thu, Sep 07, 2023 at 10:29:42PM +0300, Daniil Tatianin wrote: This series fixes an issue where the outcome of the migration qtest relies on the initial memory contents all being the same across the first 100MiB of RAM, which is a very fragile invariant. We fix this by making sure we zero the first byte of every testable page in range beforehand.What's the difference between this one and:https://lore.kernel.org/r/20230907193051.1609310-1-d-tatia...@yandex-team.ru?I think i initially forgot something and had to resend, but not sure. Anyway, disregard that one. --Peter Xu
Re: [PATCH 0/2] i386/a-b-bootblock: zero the first byte of each page on start
On Thu, Sep 07, 2023 at 10:29:42PM +0300, Daniil Tatianin wrote: > This series fixes an issue where the outcome of the migration qtest > relies on the initial memory contents all being the same across the > first 100MiB of RAM, which is a very fragile invariant. > > We fix this by making sure we zero the first byte of every testable page > in range beforehand. What's the difference between this one and: https://lore.kernel.org/r/20230907193051.1609310-1-d-tatia...@yandex-team.ru ? Thanks, -- Peter Xu
[PATCH 0/2] i386/a-b-bootblock: zero the first byte of each page on start
This series fixes an issue where the outcome of the migration qtest relies on the initial memory contents all being the same across the first 100MiB of RAM, which is a very fragile invariant. We fix this by making sure we zero the first byte of every testable page in range beforehand. Daniil Tatianin (2): i386/a-b-bootblock: factor test memory addresses out into constants i386/a-b-bootblock: zero the first byte of each page on start tests/migration/i386/a-b-bootblock.S | 18 +++--- tests/migration/i386/a-b-bootblock.h | 16 2 files changed, 23 insertions(+), 11 deletions(-) -- 2.34.1