linux-kernel - Re: virtio(-scsi) vs. chained sg_lists (was Re: [PATCH] scsi: virtio-scsi: Fix address translation failure of HighMem pages used by sg list)

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: <50103043.5050508@redhat.com>
Date:	Wed, 25 Jul 2012 19:43:31 +0200
From:	Paolo Bonzini <pbonzini@...hat.com>
To:	Boaz Harrosh <bharrosh@...asas.com>
CC:	Wang Sen <senwang@...ux.vnet.ibm.com>, linux-scsi@...r.kernel.org,
	JBottomley@...allels.com, stefanha@...ux.vnet.ibm.com,
	mc@...ux.vnet.ibm.com, linux-kernel@...r.kernel.org,
	"kvm@...r.kernel.org" <kvm@...r.kernel.org>,
	Rusty Russell <rusty@...tcorp.com.au>
Subject: Re: virtio(-scsi) vs. chained sg_lists (was Re: [PATCH] scsi: virtio-scsi:
 Fix address translation failure of HighMem pages used by sg list)

Il 25/07/2012 17:28, Boaz Harrosh ha scritto:
>> 1) what I get is a scsi_cmnd which contains an N-element scatterlist.
>>
>> 2) virtio-scsi has to build the "packet" that is passed to the hardware
>> (it does not matter that the hardware is virtual).  This packet (per
>> virtio-scsi spec) has an N+1-element scatterlist, where the first
>> element is a request descriptor (struct virtio_scsi_cmd_req), and the
>> others describe the written data.
> 
> Then "virtio-scsi spec" is crap. It overloads the meaning of
> "struct scatterlist" of the first element in an array. to be a
> "struct virtio_scsi_cmd_req".

What the holy fuck?  The first element simply _points_ to the "struct
virtio_scsi_cmd_req", just like subsequent elements point to the data.

And the protocol of the device is _not_ a struct scatterlist[].  The
virtio _API_ takes that array and converts to a series of physical
address + offset pairs.

> Since you need to change the standard to support chaining then
> it is a good time to fix this.

Perhaps it is a good time for you to read the virtio spec.  You are
making a huge confusion between the LLD->virtio interface and the
virtio->hardware interface.  I'm talking only of the former.

>> 3) virtio takes care of converting the "packet" from a scatterlist
>> (which currently must be a flat one) to the hardware representation.
>> Here a walk is inevitable, so we don't care about this walk.
> 
> "hardware representation" you mean aio or biovec, what ever the
> IO submission path uses at the host end?

No, I mean the way the virtio spec encodes the physical address + offset
pairs.

I stopped reading here.

Paolo
--
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/