linux-kernel - Re: [PATCH] ALSA: hda-intel - Disable MSI support by default

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: <4AEFD197-2F7B-4449-B072-F883F7A3721E@kernel.crashing.org>
Date:	Wed, 15 Nov 2006 11:06:15 +0100
From:	Segher Boessenkool <segher@...nel.crashing.org>
To:	Benjamin Herrenschmidt <benh@...nel.crashing.org>
Cc:	David Miller <davem@...emloft.net>, torvalds@...l.org,
	jeff@...zik.org, linux-kernel@...r.kernel.org, tiwai@...e.de
Subject: Re: [PATCH] ALSA: hda-intel - Disable MSI support by default

>> However, they have to do all the work of processing the memory
>> transation that the MSI is on the PCI bus, I don't think they  
>> would go
>> so far out of their way to reorder things even if they converted the
>> MSI packet into a PIN to the APIC, for example.
>
> That would suck and not surprise me that much in fact... Take the  
> Apple
> bridge... it's unclear at which point they actually decode the MSI and
> HT interrupts (the later are just internally converted to MSI-like
> stores) to turn them into toggling of MPIC lines,

That's all documented just fine.

> but it probably
> happens as an MMIO slave on the main xbar

Yes indeed.

> and I'm not 100% certain it
> provides ordering vs. the previous stores to memory as they are in the
> coherent domain and the MMIO is not. I just hope very much :-)

All those DMA stores get reflected to the CPUs (the north
bridge itself doesn't keep a cache directory).  All this
happens in-order.  There probably _is_ a window where the
last DMA store isn't yet globally visible but the CPU
interrupt was already signaled, but the act of trying to
access that data orders it itself :-)

> Because
> the store queue to memory can re-order on U4.

That's only the store queue to the actual memory devices,
it's outside of the coherent domain anyway.


Segher

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/