linux-kernel - Re: [PATCH] xen-pciback: Consider MSI-X enabled only when MASKALL bit is cleared

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: <1d3324da-16d2-2230-2fb4-433ef6acfc3e@suse.com>
Date:   Thu, 17 Nov 2022 14:43:01 +0100
From:   Jan Beulich <jbeulich@...e.com>
To:     Andrew Cooper <Andrew.Cooper3@...rix.com>
Cc:     Juergen Gross <jgross@...e.com>,
        Stefano Stabellini <sstabellini@...nel.org>,
        Oleksandr Tyshchenko <oleksandr_tyshchenko@...m.com>,
        "moderated list:XEN HYPERVISOR INTERFACE" 
        <xen-devel@...ts.xenproject.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Marek Marczykowski-Górecki 
        <marmarek@...isiblethingslab.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] xen-pciback: Consider MSI-X enabled only when MASKALL bit
 is cleared

On 17.11.2022 13:28, Andrew Cooper wrote:
> On 17/11/2022 11:41, Marek Marczykowski-Górecki wrote:
>> Linux enables MSI-X before disabling INTx, but keeps MSI-X masked until
>> the table is filled. Then it disables INTx just before clearing MASKALL
>> bit. Currently this approach is rejected by xen-pciback.
>> Allow setting PCI_MSIX_FLAGS_ENABLE while INTx is still enabled as long
>> as PCI_MSIX_FLAGS_MASKALL is set too.
>>
>> Fixes: 5e29500eba2a ("xen-pciback: Allow setting PCI_MSIX_FLAGS_MASKALL too")
>> Signed-off-by: Marek Marczykowski-Górecki <marmarek@...isiblethingslab.com>
> 
> The PCI spec states that devices are not permitted to use INTx when MSI
> or MSI-X is enabled.  The mask status has no legitimate bearing on irq type.
> 
> INTx_DISABLE exists as a bodge to mean "INTx not permitted even when
> neither MSI nor MSI-X are enabled", and exists because in some case,
> transiently disabling MSI is the only safe way to update the descriptor.
> 
> 
> I can believe that this change fixes a an issue, but the logic surely
> cannot be correct overall.

Question then is - what can we do without altering the sequence of steps
Linux (and likely other OSes) take? Imo Marek's proposal is the least
bad option, because everything else would be more intrusive or wouldn't
take effect for existing released kernel versions running in guests.

Jan