[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ccc138d2-fa93-43f2-9733-e461396db769@intel.com>
Date: Wed, 1 Oct 2025 07:29:14 +0200
From: Przemek Kitszel <przemyslaw.kitszel@...el.com>
To: Haotian Zhang <vulab@...as.ac.cn>, Jacob Keller
<jacob.e.keller@...el.com>, Matthew Wilcox <willy@...radead.org>
CC: Tony Nguyen <anthony.l.nguyen@...el.com>, Andrew Lunn
<andrew+netdev@...n.ch>, "David S . Miller" <davem@...emloft.net>, "Eric
Dumazet" <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo Abeni
<pabeni@...hat.com>, <intel-wired-lan@...ts.osuosl.org>,
<netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] ice: ice_adapter: release xa entry on adapter
allocation failure
On 9/30/25 03:51, Haotian Zhang wrote:
> When ice_adapter_new() fails, the reserved XArray entry created by
> xa_insert() is not released. This causes subsequent insertions at
> the same index to return -EBUSY, potentially leading to
> NULL pointer dereferences.
>
> Release the reserved entry with xa_release() when adapter allocation
> fails to prevent this issue.
>
> Fixes: 0f0023c649c7 ("ice: do not init struct ice_adapter more times than needed")
> Suggested-by: Jacob Keller <jacob.e.keller@...el.com>
> Signed-off-by: Haotian Zhang <vulab@...as.ac.cn>
>
> ---
> Changes in v2:
> - Instead of checking the return value of xa_store(), fix the real bug
> where a failed ice_adapter_new() would leave a stale entry in the
> XArray.
> - Use xa_release() to clean up the reserved entry, as suggested by
> Jacob Keller.
this is a correct improvement, but please let me propose other options,
with 2. being my favorite:
1. (just ice changes)
change the call order to be:
(note that it is under a mutex)
xa_load() // return early if another adapter exists
xa_reserve() // return early if no mem
ice_adapter_new() // return early if err
xa_store()
2. (xarray changes)
(perhaps I'm biased as the one introducing the error on error path):
change xa_insert() to return 0 or -EEXIST when used as a reserving call
(IOW: caller wanted to reserve, entry is reserved, so return should be 0
(or -EEXIST if we really want to differentiate in the callers)).
> ---
> drivers/net/ethernet/intel/ice/ice_adapter.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/intel/ice/ice_adapter.c b/drivers/net/ethernet/intel/ice/ice_adapter.c
> index b53561c34708..9eb100b11439 100644
> --- a/drivers/net/ethernet/intel/ice/ice_adapter.c
> +++ b/drivers/net/ethernet/intel/ice/ice_adapter.c
> @@ -110,8 +110,10 @@ struct ice_adapter *ice_adapter_get(struct pci_dev *pdev)
> return ERR_PTR(err);
>
> adapter = ice_adapter_new(pdev);
> - if (!adapter)
> + if (!adapter) {
> + xa_release(&ice_adapters, index);
> return ERR_PTR(-ENOMEM);
> + }
> xa_store(&ice_adapters, index, adapter, GFP_KERNEL);
> }
> return adapter;
Powered by blists - more mailing lists