lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y/aCNSlx2p62iDYk@corigine.com>
Date:   Wed, 22 Feb 2023 21:59:33 +0100
From:   Simon Horman <simon.horman@...igine.com>
To:     Sean Anderson <seanga2@...il.com>
Cc:     "David S . Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Rolf Eike Beer <eike-kernel@...tec.de>,
        kernel test robot <lkp@...el.com>,
        Dan Carpenter <error27@...il.com>
Subject: Re: [PATCH net] net: sunhme: Return an error when we are out of slots

On Wed, Feb 22, 2023 at 12:09:35PM -0500, Sean Anderson wrote:
> We only allocate enough space for four devices when the parent is a QFE. If
> we couldn't find a spot (because five devices were created for whatever
> reason), we would not return an error from probe(). Return ENODEV, which
> was what we did before.
> 
> Fixes: 96c6e9faecf1 ("sunhme: forward the error code from pci_enable_device()")

I think the hash for that commit is acb3f35f920b.

However, I also think this problem was introduced by the first hunk of
5b3dc6dda6b1 ("sunhme: Regularize probe errors").

Which is:

--- a/drivers/net/ethernet/sun/sunhme.c
+++ b/drivers/net/ethernet/sun/sunhme.c
@@ -2945,7 +2945,6 @@ static int happy_meal_pci_probe(struct pci_dev *pdev,
 	if (err)
 		goto err_out;
 	pci_set_master(pdev);
-	err = -ENODEV;
 
 	if (!strcmp(prom_name, "SUNW,qfe") || !strcmp(prom_name, "qfe")) {
 		qp = quattro_pci_find(pdev);


Which leads me to wonder if simpler fixes would be either:

1) Reverting the hunk above
2) Or, more in keeping with the rest of that patch,
   explicitly setting err before branching to err_out,
   as you your patch does, but without other logic changes.

   Something like this (*compile tested only!*:

diff --git a/drivers/net/ethernet/sun/sunhme.c b/drivers/net/ethernet/sun/sunhme.c
index 1c16548415cd..2409e7d6c29e 100644
--- a/drivers/net/ethernet/sun/sunhme.c
+++ b/drivers/net/ethernet/sun/sunhme.c
@@ -2863,8 +2863,10 @@ static int happy_meal_pci_probe(struct pci_dev *pdev,
 			if (!qp->happy_meals[qfe_slot])
 				break;
 
-		if (qfe_slot == 4)
+		if (qfe_slot == 4) {
+			err = -ENOMEM;
 			goto err_out;
+		}
 	}
 
 	dev = devm_alloc_etherdev(&pdev->dev, sizeof(struct happy_meal));

Also, I am curious why happy_meal_pci_probe() doesn't just return instaed
of branching to err_out. As err_out only returns err.  I guess there is a
reason for it. But simply returning would probably simplify error handling.
(I'm not suggesting that approach for this fix.)

> Reported-by: kernel test robot <lkp@...el.com>
> Reported-by: Dan Carpenter <error27@...il.com>
> Signed-off-by: Sean Anderson <seanga2@...il.com>
> ---
> 
>  drivers/net/ethernet/sun/sunhme.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/sun/sunhme.c b/drivers/net/ethernet/sun/sunhme.c
> index 1c16548415cd..523e26653ec8 100644
> --- a/drivers/net/ethernet/sun/sunhme.c
> +++ b/drivers/net/ethernet/sun/sunhme.c
> @@ -2861,12 +2861,13 @@ static int happy_meal_pci_probe(struct pci_dev *pdev,
>  
>  		for (qfe_slot = 0; qfe_slot < 4; qfe_slot++)
>  			if (!qp->happy_meals[qfe_slot])
> -				break;
> +				goto found_slot;
>  
> -		if (qfe_slot == 4)
> -			goto err_out;
> +		err = -ENODEV;
> +		goto err_out;
>  	}
>  
> +found_slot:
>  	dev = devm_alloc_etherdev(&pdev->dev, sizeof(struct happy_meal));
>  	if (!dev) {
>  		err = -ENOMEM;
> -- 
> 2.37.1
> 

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ