netdev - Re: [PATCH net] nfp: do not use driver

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] [day] [month] [year] [list]

Message-ID: <20220401102230.2caab128@kernel.org>
Date:   Fri, 1 Apr 2022 10:22:30 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Simon Horman <simon.horman@...igine.com>
Cc:     David Miller <davem@...emloft.net>, netdev@...r.kernel.org,
        oss-drivers@...igine.com,
        Niklas Söderlund 
        <niklas.soderlund@...igine.com>,
        Danie du Toit <danie.dutoit@...igine.com>
Subject: Re: [PATCH net] nfp: do not use driver_data to index device info

On Fri,  1 Apr 2022 13:19:36 +0200 Simon Horman wrote:
> From: Niklas Söderlund <niklas.soderlund@...igine.com>
> 
> When adding support for multiple chips the struct pci_device_id
> driver_data field was used to hold a index to lookup chip device
> specific information from a table. This works but creates a regressions
> for users who uses /sys/bus/pci/drivers/nfp_netvf/new_id.
> 
> For example, before the change writing "19ee 6003" to new_id was
> sufficient 

Can you explain the use case? I think this worked somewhat
coincidentally. If we had entries that matched subvendor = 0
subdevice = 0 it'd fail with EEXIST.

> but after one needs to write enough fields to be able to also
> match on the driver_data field, "19ee 6003 19ee ffffffff ffffffff 0 1".
> 
> The usage of driver_data field was only a convenience and in the belief
> the driver_data field was private to the driver and not exposed in
> anyway to users. Changing the device info lookup to a function that
> translates from struct pci_device_id device field instead works just as
> well and removes the user facing regression.

I think you're trading a coincidental "feature" while breaking what
new_id is actually supposed to be used for. Which is adding IDs.
nfp_get_dev_info() you add only recognizes existing IDs.

> As a bonus the enum and table with lookup information can be moved out
> from a shared header file to the only file where it's used.
> 
> Reported-by: Danie du Toit <danie.dutoit@...igine.com>
> Fixes: e900db704c8512bc ("nfp: parametrize QCP offset/size using dev_info")
> Signed-off-by: Niklas Söderlund <niklas.soderlund@...igine.com>