Codestin Search App

poliglots · 2026-08-22T09:01:06Z

When hugepages=on but reserve isn't explicitly set, default reserve to on. This ensures the kernel reserves physical RAM for huge pages at boot time, exposing memory shortages early as a configuration error rather than later via a random SIGBUS mid-flight.

prefault=on takes precedence and is not affected, since prefaulting already implies reservation. Users can still explicitly set reserve=off.

Fixes #8742

rbradford

Thanks but this is wrong. We talk about the CLI as a shorthand but we also need to cover the API boot. Also we need to be clever and use Option type for compatibility.

poliglots · 2026-08-22T09:29:38Z

Thanks @rbradford for the feedback, let me try to cover them. will update you.

poliglots · 2026-08-22T12:39:01Z

Hi @rbradford ,
I have tried to cover the API part. Please check if I am in right direction.
Thanks.

rbradford

Thank you for taking my feedback onboard.

rbradford · 2026-08-24T08:21:41Z

+impl MemoryZoneConfig {
+    pub fn reserve(&self) -> bool {
+        self.reserve.unwrap_or(self.hugepages && !self.prefault)
+    }
+}
+


Not quite right. Apply the policy only in MemoryManager

rbradford · 2026-08-24T08:21:52Z

+impl MemoryConfig {
+    pub fn reserve(&self) -> bool {
+        self.reserve.unwrap_or(self.hugepages && !self.prefault)
+    }
+}
+


rbradford · 2026-08-24T08:22:20Z

                    hugepage_size: None,
                    prefault: false,
-                    reserve: false,
+                    reserve: None,


Should be rolled into the commit that broke the build. Every commit needs to be buildable.

poliglots · 2026-08-24T08:30:25Z

Thanks. Will work on it and update you.

poliglots · 2026-08-24T13:22:16Z

Hi, I have made the changes as suggested by you. Please take a look. Thanks.

rbradford

Please audit your commits. You shouldn't add and then remove.

poliglots · 2026-08-24T14:37:54Z

Sorry for the mess. Will correct and update by tomorrow. Thanks for feedback.

poliglots · 2026-08-25T08:38:41Z

Hi Rob,

Thanks for all the feedback. I've squashed all 4 commits into one and moved the reserve default policy so it's applied only in MemoryManager, with the config structs keeping Option for API compatibility.

Apologies for the messy history — appreciate your patience.

Could you please take another look?

rbradford

You should still keep the separate commit for making reserve an Option<bool>

rbradford · 2026-08-25T08:43:00Z

                hugepage_size: Some(2 << 20),
                size: 1 << 30,
                hugepages: true,
+                // reserve is None; the effective default is computed in MemoryManager


You can drop this comment

rbradford · 2026-08-25T08:50:20Z

                    region_size as usize,
                    prefault.unwrap_or(zone.prefault),
-                    zone.reserve,
+                    zone.reserve.unwrap_or(zone.hugepages && !zone.prefault),


Do you actually need the !zone.prefault ? How does this affect the flags in create_ram_region_raw() does it even matter?

rbradford · 2026-08-25T08:51:30Z

I would look into git rebase -i if you don't know it. It will help you create an appropriately structured set of commits,

poliglots · 2026-08-25T10:37:20Z

Thanks, will update by tomorrow.

poliglots · 2026-08-25T11:20:03Z

You should still keep the separate commit for making reserve an Option<bool>

Hi @rbradford ,

Need your little guidance here. I am planning to have two commits

First commit will the copy of "8c8fdc0"
Second commit will the diff of current commit "7328cc6" and "8c8fdc0".

I have Question:

If I make separate commit 8c8fdc0 i.e. Use Option for reserve and add .reserve() method , then the "impl MemoryConfig" which I defined in config.rs will be get removed in 2nd commit. Then will it look like I removed again after adding. or its ok since we do it because of review comments.
In the Second commit , I will incorporate the latest feedback i.e. removal of !zone.prefault from "memory_manager.rs" and removal of comment from "config.rs" along with policy application in memory manager.

Do you think this would be correct approach.

Thanks.

rbradford · 2026-08-25T13:03:10Z

I

You should still keep the separate commit for making reserve an Option<bool>

Hi @rbradford ,

Need your little guidance here. I am planning to have two commits

First commit will the copy of "8c8fdc0"

Second commit will the diff of current commit "7328cc6" and "8c8fdc0".

I have Question:

If I make separate commit 8c8fdc0 i.e. Use Option for reserve and add .reserve() method , then the "impl MemoryConfig" which I defined in config.rs will be get removed in 2nd commit. Then will it look like I removed again after adding. or its ok since we do it because of review comments.

In the Second commit , I will incorporate the latest feedback i.e. removal of !zone.prefault from "memory_manager.rs" and removal of comment from "config.rs" along with policy application in memory manager.

I don't know why you think you need to reference these old commits. Instead just split the commit you already had? Why would feel you need to add the helper when you can just to .unwrap_or(false).

I took 5 minutes and updated the PR with the appropriate structure. Hope you don't mind. I think were were going around in circles over something which should be very simple.

Make specifying this optional rather than a default of false so as to allow differentiating from intentionally included and false vs ambivalent. This prepares for enabling reserve when hugepages is enabled when the user does not intentionally disable reserve. See: cloud-hypervisor#8742 Signed-off-by: Shatrughan Rai <[email protected]> Signed-off-by: Rob Bradford <[email protected]>

rbradford · 2026-08-25T13:12:52Z

I had to rebase and address some issues due to #8581 landing.

Don't map with MADV_NORESERVE when hugepages are enabled unless the user has explicitly disabled reserving. Fixes: cloud-hypervisor#8742 Signed-off-by: Shatrughan Rai <[email protected]> Signed-off-by: Rob Bradford <[email protected]>

poliglots · 2026-08-25T13:41:01Z

Hi Rob,

I completely agree that we were going in circles. Thanks for putting the commit "a35e6ba" and "39683f1"

Your feedback will help me to improve for future PRs.

Thanks.

rbradford · 2026-08-25T13:45:50Z

Hi Rob,

I completely agree that we were going in circles. Thanks for putting the commit "a35e6ba" and "39683f1"

Your feedback will help me to improve for future PRs.

Thanks.

Good to hear. Hope you can understand why this split makes sense. In this case: 1. prep (and yes that meant adding .unwrap_of(false) that would be immediately replaced) 2. apply policy standalone so it's easy to understand

poliglots · 2026-08-26T07:24:59Z

Good to hear. Hope you can understand why this split makes sense. In this case: 1. prep (and yes that meant adding .unwrap_of(false) that would be immediately replaced) 2. apply policy standalone so it's easy to understand

Thanks for the explanation, The two-step split makes complete sense. Thanks.

poliglots requested a review from a team as a code owner August 22, 2026 09:01

rbradford requested changes Aug 22, 2026

View reviewed changes

poliglots force-pushed the main branch from fbdff2c to 3dc0a2b Compare August 22, 2026 11:58

poliglots requested a review from rbradford August 24, 2026 06:08

rbradford requested changes Aug 24, 2026

View reviewed changes

poliglots force-pushed the main branch from b00ab10 to 8c8fdc0 Compare August 24, 2026 08:57

poliglots requested a review from rbradford August 24, 2026 13:22

rbradford requested changes Aug 24, 2026

View reviewed changes

poliglots force-pushed the main branch from 7f1e498 to 7328cc6 Compare August 25, 2026 06:48

poliglots requested a review from rbradford August 25, 2026 08:40

rbradford reviewed Aug 25, 2026

View reviewed changes

rbradford force-pushed the main branch from 7328cc6 to 91fbf4a Compare August 25, 2026 13:00

rbradford force-pushed the main branch from 91fbf4a to a35e6ba Compare August 25, 2026 13:12

rbradford force-pushed the main branch from a35e6ba to ae66d59 Compare August 25, 2026 13:39

rbradford enabled auto-merge August 25, 2026 13:46

rbradford approved these changes Aug 26, 2026

View reviewed changes

rbradford added this pull request to the merge queue Aug 26, 2026

Merged via the queue into cloud-hypervisor:main with commit 1a576b4 Aug 26, 2026
42 checks passed

Uh oh!

Conversation

poliglots commented Aug 22, 2026

Uh oh!

rbradford left a comment

Choose a reason for hiding this comment

Uh oh!

poliglots commented Aug 22, 2026

Uh oh!

poliglots commented Aug 22, 2026

Uh oh!

rbradford left a comment

Choose a reason for hiding this comment

Uh oh!

rbradford Aug 24, 2026

Choose a reason for hiding this comment

Uh oh!

rbradford Aug 24, 2026

Choose a reason for hiding this comment

Uh oh!

rbradford Aug 24, 2026

Choose a reason for hiding this comment

Uh oh!

poliglots commented Aug 24, 2026

Uh oh!

poliglots commented Aug 24, 2026

Uh oh!

rbradford left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

poliglots commented Aug 24, 2026

Uh oh!

poliglots commented Aug 25, 2026

Uh oh!

rbradford left a comment

Choose a reason for hiding this comment

Uh oh!

rbradford Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

rbradford Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

rbradford commented Aug 25, 2026

Uh oh!

poliglots commented Aug 25, 2026

Uh oh!

poliglots commented Aug 25, 2026

Uh oh!

rbradford commented Aug 25, 2026

Uh oh!

rbradford commented Aug 25, 2026

Uh oh!

poliglots commented Aug 25, 2026

Uh oh!

rbradford commented Aug 25, 2026

Uh oh!

poliglots commented Aug 26, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

rbradford left a comment •

edited

Loading