qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] Using PCI config space to indicate config location


From: Rusty Russell
Subject: Re: [Qemu-devel] Using PCI config space to indicate config location
Date: Thu, 11 Oct 2012 11:48:22 +1030
User-agent: Notmuch/0.13.2 (http://notmuchmail.org) Emacs/23.3.1 (i686-pc-linux-gnu)

"Michael S. Tsirkin" <address@hidden> writes:
> On Mon, Oct 08, 2012 at 12:51:25PM +1030, Rusty Russell wrote:
>> Note before anyone gets confused; we were talking about using the PCI
>> config space to indicate what BAR(s) the virtio stuff is in.  An
>> alternative would be to simply specify a new layout format in BAR1.
>
> One problem we are still left with is this: device specific
> config accesses are still non atomic.
> This is a problem for multibyte fields such as MAC address
> where MAC could change while we are accessing it.

It's also a problem for related fields, eg. console width and height, or
disk geometry.

> I was thinking about some backwards compatible way to solve this, but if
> we are willing to break compatiblity or use some mode switch, how about
> we give up on virtio config space completely, and do everything besides
> IO and ISR through guest memory?

I think there's still a benefit in the simple publishing of information:
I don't really want to add a control queue for the console.  But
inevitably, once-static information can change in later versions, and
it's horrible to have config information plus a bit that says "don't use
this, use the control queue".

Here's a table from a quick audit:

Driver          Config       Device changes    Driver writes... after init?
net             Y            Y                 N                N
block           Y            Y                 Y                Y
console         Y            Y                 N                N
rng             N            N                 N                N
balloon         Y            Y                 Y                Y
scsi            Y            N                 Y                N
9p              Y            N                 N                N

For config space reads, I suggest the driver publish a generation count.
For writes, the standard seems to be a commit latch.  We could abuse the
generation count for this: the driver writes to it to commit config
changes.

ie:
/* Fields in VIRTIO_PCI_CAP_COMMON_CFG: */
struct virtio_pci_common_cfg {
        /* About the whole device. */
        __le32 device_feature_select;   /* read-write */
        __le32 device_feature;          /* read-only */
        __le32 guest_feature_select;    /* read-write */
        __le32 guest_feature;           /* read-only */
        __le32 config_gen_and_latch;    /* read-write */
        __le16 msix_config;             /* read-write */
        __u8 device_status;             /* read-write */
        __u8 unused;

        /* About a specific virtqueue. */
        __le16 queue_select;    /* read-write */
        __le16 queue_align;     /* read-write, power of 2. */
        __le16 queue_size;      /* read-write, power of 2. */
        __le16 queue_msix_vector;/* read-write */
        __le64 queue_address;   /* read-write: 0xFFFFFFFFFFFFFFFF == DNE. */
};

Thoughts?
Rusty.
PS.  Let's make all the virtio-device config LE, too...



reply via email to

[Prev in Thread] Current Thread [Next in Thread]