Re: [RFC PATCH] Fix race in live migration failure path

2025-01-23 Thread Peter Xu
On Thu, Jan 23, 2025 at 06:05:28PM +, Shivam Kumar wrote: > > > > On 23 Jan 2025, at 9:57 PM, Peter Xu wrote: > > > > !---| > > CAUTION: External Email > > > > |

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-23 Thread Shivam Kumar
> On 23 Jan 2025, at 9:57 PM, Peter Xu wrote: > > !---| > CAUTION: External Email > > |---! > > On Thu, Jan 23, 2025 at 09:53:16AM +, Shivam Kumar wrote: >> >>

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-23 Thread Peter Xu
On Thu, Jan 23, 2025 at 09:53:16AM +, Shivam Kumar wrote: > > > > On 22 Jan 2025, at 10:10 PM, Peter Xu wrote: > > > > !---| > > CAUTION: External Email > > > > |---

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-23 Thread Shivam Kumar
> On 22 Jan 2025, at 10:10 PM, Peter Xu wrote: > > !---| > CAUTION: External Email > > |---! > > Hi, Shivam, > > On Wed, Jan 22, 2025 at 10:54:17AM +, Shivam K

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-22 Thread Peter Xu
Hi, Shivam, On Wed, Jan 22, 2025 at 10:54:17AM +, Shivam Kumar wrote: > There is one place where we set the migration status to FAILED but we don’t > set > s->error, i.e. in save_snapshot workflow, please check qemu_savevm_state but > not sure setting s->error in this case is possible (or req

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-22 Thread Shivam Kumar
On 13 Jan 2025, at 9:59 PM, Peter Xu wrote: !---| CAUTION: External Email |---! On Fri, Jan 10, 2025 at 10:09:38AM -0300, Fabiano Rosas wrote: Shivam Kumar writes:

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-13 Thread Peter Xu
On Fri, Jan 10, 2025 at 10:09:38AM -0300, Fabiano Rosas wrote: > Shivam Kumar writes: > > > Even if a live migration fails due to some reason, migration status > > should not be set to MIGRATION_STATUS_FAILED until migrate fd cleanup > > is done, else the client can trigger another instance of mi

Re: [RFC PATCH] Fix race in live migration failure path

2025-01-10 Thread Fabiano Rosas
Shivam Kumar writes: > Even if a live migration fails due to some reason, migration status > should not be set to MIGRATION_STATUS_FAILED until migrate fd cleanup > is done, else the client can trigger another instance of migration > before the cleanup is complete (as it would assume no migration

[RFC PATCH] Fix race in live migration failure path

2025-01-10 Thread Shivam Kumar
Even if a live migration fails due to some reason, migration status should not be set to MIGRATION_STATUS_FAILED until migrate fd cleanup is done, else the client can trigger another instance of migration before the cleanup is complete (as it would assume no migration is active) or reset migration