Skip to contentSkip to conversation

rds: mark snapshot pages dirty in rds_info_getsockopt()

PATCH Applied Breno Leitao 3 participants 5 messages Jun 8 – Jun 10

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 rds: mark snapshot pages dirty in rds_info_getsockopt()

rds_info_getsockopt() pins the destination user pages with FOLL_WRITE and the RDS_INFO_* producers memcpy the snapshot into them through kmap_atomic(). Because that copy goes through the kernel direct map, the dirty bit on the user PTE is never set, so unpin_user_pages() releases the pages without marking them dirty. A file-backed destination page can then be reclaimed without writeback, silently discarding the copied data.

Use unpin_user_pages_dirty_lock() with make_dirty=true so the modified pages are marked dirty before they are unpinned.

Fixes a8c879a7ee98 ("RDS: Info and stats")
---
 net/rds/info.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/rds/info.c +1 -1
index f1b29994934a..17061f6ff74e 100644
@@ -235,7 +235,7 @@ int rds_info_getsockopt(struct socket *sock, int optname, char __user *optval,
 
 out:
 	if (pages)
-		unpin_user_pages(pages, nr_pages);
+		unpin_user_pages_dirty_lock(pages, nr_pages, true);
 	kfree(pages);
 
 	return ret;

---
base-commit: 9772589b57e44aedc240211c5c3f7a684a034d3a
change-id: 20260608-rds_fix-92d6c7f04fe2

Best regards,

Signature · 1 line
Breno Leitao <[email protected]>
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 7 lines · Breno Leitao

rds_info_getsockopt() pins the destination user pages with FOLL_WRITE and the RDS_INFO_* producers memcpy the snapshot into them through kmap_atomic(). Because that copy goes through the kernel direct map, the dirty bit on the user PTE is never set, so unpin_user_pages() releases the pages without marking them dirty. A file-backed destination page can then be reclaimed without writeback, silently discarding the copied data.

Quoted 5 lines

Use unpin_user_pages_dirty_lock() with make_dirty=true so the modified pages are marked dirty before they are unpinned.

Fixes a8c879a7ee98 ("RDS: Info and stats")

Hi Breno,

Thanks for adding the Fixes tag. One thing though: now that this is standalone, it collides with "[PATCH net-next v3 2/2] rds: convert to getsockopt_iter" since both rewrite the same unpin in the out: block of rds_info_getsockopt().

Easiest on everyone is to just keep it folded into the net-next series as a three-patch set, rather than splitting the fix out to net.

With that, you can add my:

Thanks!
Allison

Quoted 23 lines
---
 net/rds/info.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/rds/info.c +1 -2
index f1b29994934a..17061f6ff74e 100644
@@ -235,7 +235,7 @@ int rds_info_getsockopt(struct socket *sock, int optname, char __user *optval,
 
 out:
 	if (pages)
-		unpin_user_pages(pages, nr_pages);
+		unpin_user_pages_dirty_lock(pages, nr_pages, true);
 	kfree(pages);
 
 	return ret;
 
---

base-commit: 9772589b57e44aedc240211c5c3f7a684a034d3a
change-id: 20260608-rds_fix-92d6c7f04fe2

Best regards,

#2 lore raw
Quoted 16 lines · Allison Henderson
Quoted 7 lines · Breno Leitao

rds_info_getsockopt() pins the destination user pages with FOLL_WRITE and the RDS_INFO_* producers memcpy the snapshot into them through kmap_atomic(). Because that copy goes through the kernel direct map, the dirty bit on the user PTE is never set, so unpin_user_pages() releases the pages without marking them dirty. A file-backed destination page can then be reclaimed without writeback, silently discarding the copied data.

Quoted 5 lines

Use unpin_user_pages_dirty_lock() with make_dirty=true so the modified pages are marked dirty before they are unpinned.

Fixes a8c879a7ee98 ("RDS: Info and stats")

Hi Breno,

Quoted 7 lines

Thanks for adding the Fixes tag. One thing though: now that this is standalone, it collides with "[PATCH net-next v3 2/2] rds: convert to getsockopt_iter" since both rewrite the same unpin in the out: block of rds_info_getsockopt().

Easiest on everyone is to just keep it folded into the net-next series as a three-patch set, rather than splitting the fix out to net.

I got from the maintainers that they want to continue to split the fixes into net and non fixes (features) into net-next.

"So just prepare for net with Fixes tags and we'll route the patches accordingly." -- Jakub

https://lore.kernel.org/all/[email protected]/

Thanks for the review,
--breno

#3 lore raw
Quoted 28 lines · Breno Leitao
Quoted 16 lines · Allison Henderson
Quoted 7 lines · Breno Leitao

rds_info_getsockopt() pins the destination user pages with FOLL_WRITE and the RDS_INFO_* producers memcpy the snapshot into them through kmap_atomic(). Because that copy goes through the kernel direct map, the dirty bit on the user PTE is never set, so unpin_user_pages() releases the pages without marking them dirty. A file-backed destination page can then be reclaimed without writeback, silently discarding the copied data.

Quoted 5 lines

Use unpin_user_pages_dirty_lock() with make_dirty=true so the modified pages are marked dirty before they are unpinned.

Fixes a8c879a7ee98 ("RDS: Info and stats")

Hi Breno,

Quoted 7 lines

Thanks for adding the Fixes tag. One thing though: now that this is standalone, it collides with "[PATCH net-next v3 2/2] rds: convert to getsockopt_iter" since both rewrite the same unpin in the out: block of rds_info_getsockopt().

Easiest on everyone is to just keep it folded into the net-next series as a three-patch set, rather than splitting the fix out to net.

I got from the maintainers that they want to continue to split the fixes into net and non fixes (features) into net-next.

Quoted 7 lines

"So just prepare for net with Fixes tags and we'll route the patches accordingly." -- Jakub

https://lore.kernel.org/all/[email protected]/

Thanks for the review,
--breno

I see, ok it's fine to apply my rvb then.

Thanks for working on this,
Allison

#4 lore raw
In reply to Breno Leitao
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 8 lines · On Mon, 08 Jun 2026 02:32:05 -0700 you wrote:

rds_info_getsockopt() pins the destination user pages with FOLL_WRITE and the RDS_INFO_* producers memcpy the snapshot into them through kmap_atomic(). Because that copy goes through the kernel direct map, the dirty bit on the user PTE is never set, so unpin_user_pages() releases the pages without marking them dirty. A file-backed destination page can then be reclaimed without writeback, silently discarding the copied data.

[...]

Here is the summary with links:
- [net] rds: mark snapshot pages dirty in rds_info_getsockopt()

    https://git.kernel.org/netdev/net/c/512db8267b73

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html