From 306baacdd420a62903dd496624af10c6c96669bb Mon Sep 17 00:00:00 2001 From: rusty Date: Thu, 31 Oct 2013 05:28:01 +0000 Subject: Configuration (read) atomicity. Aka issue VIRTIO-35. This is solved per transport: 1) PCI: use the 8 bit reserved field. Assume that if you really change that fast, you'll do it lazily on config space read. 2) MMIO (already solved by v2 update) 3) CCW: no transport changes. They always read/write the entire thing. This just shows that Cornelia is smarter than I am. Signed-off-by: Rusty Russell git-svn-id: https://tools.oasis-open.org/version-control/svn/virtio@74 0c8fb4dd-22a2-4bb5-bc14-6c75a5f43652 --- virtio-v1.0-wd01-part1-specification.txt | 53 ++++++++++++++++++++++++++++---- 1 file changed, 47 insertions(+), 6 deletions(-) diff --git a/virtio-v1.0-wd01-part1-specification.txt b/virtio-v1.0-wd01-part1-specification.txt index ae79444..908d631 100644 --- a/virtio-v1.0-wd01-part1-specification.txt +++ b/virtio-v1.0-wd01-part1-specification.txt @@ -208,11 +208,36 @@ Interface' in the section title. ------------------------- Configuration space is generally used for rarely-changing or -initialization-time parameters. +initialization-time parameters. Drivers must not assume reads from +fields greater than 32 bits wide are atomic, nor or reads from +multiple fields. + +Each transport provides a generation count for the configuration +space, which must change whenever there is a possibility that two +accesses to the configuration space can see different versions of that +space. + +Thus drivers should read configuration space fields like so: + + u32 before, after; + do { + before = get_config_generation(device); + // read config entry/entries. + after = get_config_generation(device); + } while (after != before); Note that this space is generally the guest's native endian, rather than PCI's little-endian. +2.1.3.1. Legacy Interface: Configuration Space +------------------------- + +Legacy devices did not have a configuration generation field, thus are +susceptible to race conditions if configuration is updated. This +effects the block capacity and network mac fields; best practice is to +read these fields multiple times until two reads generate a consistent +result. + 2.1.4. Virtqueues ---------------- @@ -801,7 +826,7 @@ struct virtio_pci_common_cfg { __le16 msix_config; /* read-write */ __le16 num_queues; /* read-only */ __u8 device_status; /* read-write */ - __u8 unused1; + __u8 config_generation; /* read-only */ /* About a specific virtqueue. */ __le16 queue_select; /* read-write */ @@ -853,6 +878,14 @@ device_status Device Status field. Writing 0 into this field resets the device. +config_generation + + Configuration atomicity value. Changes every time the + configuration noticeably changes. This means the device may + only change the value after a configuration read operation, + but it must change if there is any risk of a device seeing an + inconsistent configuration state. + queue_select Queue Select. Selects which virtqueue do other fields refer to. @@ -975,6 +1008,9 @@ Legacy Interface. When used through the Legacy Interface, Transitional Devices must assume that Feature Bits 32 to 63 are not acknowledged by Driver. +As legacy devices had no configuration generation field, see "2.1.3.1. +Legacy Interface: Configuration Space" for workarounds. + 2.3.1.3. PCI-specific Initialization And Device Operation -------------------------------------------------------- @@ -1859,7 +1895,9 @@ For changing configuration information, the guest may use CCW_CMD_WRITE_CONF, specifying the guest memory for the host to read from. -In both cases, the complete configuration space is transmitted. +In both cases, the complete configuration space is transmitted. This +allows the guest to compare the new configuration space with the old +version, and keep a generation count internally whenever it changes. 2.3.3.2.5. Setting Up Indicators -------------------------------- @@ -3792,12 +3830,15 @@ transmit output. --------------------------------------- Configuration space should only be used for initialization-time -parameters. It is a limited resource with no synchronization, so for -most uses it is better to use a virtqueue to update configuration -information (the network device does this for filtering, +parameters. It is a limited resource with no synchronization between +writable fields, so for most uses it is better to use a virtqueue to update +configuration information (the network device does this for filtering, otherwise the table in the config space could potentially be very large). +Devices must not assume that configuration fields over 32 bits wide +are atomically writable. + 2.7.3. What Device Number? -------------------------- -- cgit v1.2.3