[PATCH] liveupdate: check for duplicate session before growing block set

[PATCH] liveupdate: check for duplicate session before growing block set

[Christian Aurich Zanettini Martins]

luo_session_insert() grows the outgoing block set before checking for a
duplicate session name. On -EEXIST the new block stays in the set.

If the session count is an exact multiple of the per-block capacity (51
with 4K pages), that block stays empty. It is serialized with count 0,
which kho_block_set_restore() rejects in the next kernel:

  luo: Block contains invalid entry count: 0
  Kernel panic - not syncing: The incoming tree failed to initialize properly [-EINVAL], disabling live update

Check for the duplicate before growing the block set, as
luo_preserve_file() does with the token.

Fixes: 2a441a14c2c0 ("liveupdate: Remove limit on the number of sessions")
Cc: stable@vger.kernel.org
Signed-off-by: Christian Aurich Zanettini Martins <christian.aurichzm@gmail.com>
---

Notes:
    Tested in QEMU: create N sessions, retry "sess-0" to get -EEXIST, then
    kexec and retrieve them. Unpatched, the second kernel panics for N=51
    and N=102; patched, N=1, 51, 52, 102 and 153 all come back.

    Based on v7.4-rc1.

 kernel/liveupdate/luo_session.c | 21 +++++++++++----------
 1 file changed, 11 insertions(+), 10 deletions(-)

diff --git a/kernel/liveupdate/luo_session.c b/kernel/liveupdate/luo_session.c
index f48e9a4185f9..74bc66929f4b 100644
--- a/kernel/liveupdate/luo_session.c
+++ b/kernel/liveupdate/luo_session.c
@@ -176,16 +176,6 @@ static int luo_session_insert(struct luo_session_header *sh,
 
 	guard(rwsem_write)(&sh->rwsem);
 
-	/*
-	 * For outgoing we should make sure there is room in serialization array
-	 * for new session.
-	 */
-	if (sh == &luo_session_global.outgoing) {
-		err = kho_block_set_grow(&sh->block_set, sh->count + 1);
-		if (err)
-			return err;
-	}
-
 	/*
 	 * For small number of sessions this loop won't hurt performance
 	 * but if we ever start using a lot of sessions, this might
@@ -196,6 +186,17 @@ static int luo_session_insert(struct luo_session_header *sh,
 		if (!strncmp(it->name, session->name, sizeof(it->name)))
 			return -EEXIST;
 	}
+
+	/*
+	 * For outgoing we should make sure there is room in serialization array
+	 * for new session.
+	 */
+	if (sh == &luo_session_global.outgoing) {
+		err = kho_block_set_grow(&sh->block_set, sh->count + 1);
+		if (err)
+			return err;
+	}
+
 	list_add_tail(&session->list, &sh->list);
 	sh->count++;
 

base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4
-- 
2.43.0

Re: [PATCH] liveupdate: check for duplicate session before growing block set

[Pratyush Yadav]

On Tue, Nov 03 2026, Christian Aurich Zanettini Martins wrote:

> luo_session_insert() grows the outgoing block set before checking for a
> duplicate session name. On -EEXIST the new block stays in the set.
>
> If the session count is an exact multiple of the per-block capacity (51
> with 4K pages), that block stays empty. It is serialized with count 0,
> which kho_block_set_restore() rejects in the next kernel:
>
>   luo: Block contains invalid entry count: 0
>   Kernel panic - not syncing: The incoming tree failed to initialize properly [-EINVAL], disabling live update
>
> Check for the duplicate before growing the block set, as
> luo_preserve_file() does with the token.
>
> Fixes: 2a441a14c2c0 ("liveupdate: Remove limit on the number of sessions")
> Cc: stable@vger.kernel.org
> Signed-off-by: Christian Aurich Zanettini Martins <christian.aurichzm@gmail.com>

Reviewed-by: Pratyush Yadav (Google) <pratyush@kernel.org>

[...]

-- 
Regards,
Pratyush Yadav

Re: [PATCH] liveupdate: check for duplicate session before growing block set

[Mike Rapoport]

From: Mike Rapoport (Microsoft) <rppt@kernel.org>

On Tue, 03 Nov 2026 12:39:28 -0300, Christian Aurich Zanettini Martins wrote:
> luo_session_insert() grows the outgoing block set before checking for a
> duplicate session name. On -EEXIST the new block stays in the set.
> 
> If the session count is an exact multiple of the per-block capacity (51
> with 4K pages), that block stays empty. It is serialized with count 0,
> which kho_block_set_restore() rejects in the next kernel:
> 
> [...]

Applied to next branch of liveupdate/linux.git tree, thanks!

[1/1] liveupdate: check for duplicate session before growing block set
      commit: 7d4282ec2d3bfd5e7959bbd2663936a29c45f609

tree: https://git.kernel.org/pub/scm/linux/kernel/git/liveupdate/linux
branch: next

--
Sincerely yours,
Mike.

Re: [PATCH] liveupdate: check for duplicate session before growing block set

[Pasha Tatashin]

On 11-03 12:39, Christian Aurich Zanettini Martins wrote:
> luo_session_insert() grows the outgoing block set before checking for a
> duplicate session name. On -EEXIST the new block stays in the set.
> 
> If the session count is an exact multiple of the per-block capacity (51
> with 4K pages), that block stays empty. It is serialized with count 0,
> which kho_block_set_restore() rejects in the next kernel:
> 
>   luo: Block contains invalid entry count: 0
>   Kernel panic - not syncing: The incoming tree failed to initialize properly [-EINVAL], disabling live update
> 
> Check for the duplicate before growing the block set, as
> luo_preserve_file() does with the token.
> 
> Fixes: 2a441a14c2c0 ("liveupdate: Remove limit on the number of sessions")
> Cc: stable@vger.kernel.org
> Signed-off-by: Christian Aurich Zanettini Martins <christian.aurichzm@gmail.com>
> ---
> 
> Notes:
>     Tested in QEMU: create N sessions, retry "sess-0" to get -EEXIST, then
>     kexec and retrieve them. Unpatched, the second kernel panics for N=51
>     and N=102; patched, N=1, 51, 52, 102 and 153 all come back.
> 
>     Based on v7.4-rc1.
> 
>  kernel/liveupdate/luo_session.c | 21 +++++++++++----------
>  1 file changed, 11 insertions(+), 10 deletions(-)
> 
> diff --git a/kernel/liveupdate/luo_session.c b/kernel/liveupdate/luo_session.c
> index f48e9a4185f9..74bc66929f4b 100644
> --- a/kernel/liveupdate/luo_session.c
> +++ b/kernel/liveupdate/luo_session.c
> @@ -176,16 +176,6 @@ static int luo_session_insert(struct luo_session_header *sh,
>  
>  	guard(rwsem_write)(&sh->rwsem);
>  
> -	/*
> -	 * For outgoing we should make sure there is room in serialization array
> -	 * for new session.
> -	 */
> -	if (sh == &luo_session_global.outgoing) {
> -		err = kho_block_set_grow(&sh->block_set, sh->count + 1);
> -		if (err)
> -			return err;
> -	}
> -
>  	/*
>  	 * For small number of sessions this loop won't hurt performance
>  	 * but if we ever start using a lot of sessions, this might
> @@ -196,6 +186,17 @@ static int luo_session_insert(struct luo_session_header *sh,
>  		if (!strncmp(it->name, session->name, sizeof(it->name)))
>  			return -EEXIST;
>  	}
> +
> +	/*
> +	 * For outgoing we should make sure there is room in serialization array
> +	 * for new session.
> +	 */
> +	if (sh == &luo_session_global.outgoing) {
> +		err = kho_block_set_grow(&sh->block_set, sh->count + 1);
> +		if (err)
> +			return err;
> +	}
> +

Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>