audk/OvmfPkg/Library/VirtioLib/VirtioLib.c

553 lines
18 KiB
C
Raw Normal View History

/** @file
Utility functions used by virtio device drivers.
OvmfPkg: VirtioFlush(): return the number of bytes written by the host VirtioLib provides an API for simple, synchronous (request/response-style) virtio communication. The guest driver builds one descriptor chain, link for link, with VirtioPrepare() and VirtioAppendDesc(), then submits the chain, and awaits the processing, with VirtioFlush(). The descriptor chain is always built at the beginning of the descriptor area, with the head descriptor having descriptor index 0. In order to submit the descriptor chain to the host, the guest always pushes a new "available element" to the Available Ring, in genuine queue-like fashion, with the new element referencing the head descriptor (which always has index 0, see above). In turn, after processing, the host always pushes a new "used element" to the Used Ring, in genuine queue-like fashion, with the new element referencing the head descriptor of the chain that was just processed. The same element also reports the number of bytes that the host wrote, consecutively across the host-writeable buffers that were linked by the descriptors. (See "OvmfPkg/VirtioNetDxe/TechNotes.txt" for a diagram about the descriptor area and the rings.) Because at most one descriptor chain can be in flight with VirtioLib at any time, - the Available Ring and the Used Ring proceed in lock-step, - and the head descriptor that the new "available" and "used" elements can ever reference has index 0. Based on the above, we can modify VirtioFlush() to return the number of bytes written by the host across the descriptor chain. The virtio-block and virtio-scsi drivers don't care (they have other ways to parse the data produced by the host), while the virtio-net driver doesn't use VirtioFlush() at all (it employs VirtioLib only to set up its rings). However, the virtio entropy device, to be covered in the upcoming patches, reports the amount of randomness produced by the host only through this quantity. Cc: Jordan Justen <jordan.l.justen@intel.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-02-19 18:15:07 +01:00
Copyright (C) 2012-2016, Red Hat, Inc.
OvmfPkg: Make the VirtIo devices use the new VIRTIO_DEVICE_PROTOCOL This change replaces the accesses to the PCI bus from the Block, Scsi and Net drivers by the use of the new VIRTIO_DEVICE_PROTOCOL protocol that abstracts the transport layer. It means these drivers can be used on PCI and MMIO transport layer. Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Olivier Martin <olivier.martin@arm.com> v5: - VirtioFlush(): update comment block in VirtioLib.[hc]; error code is propagated from VirtIo->SetQueueNotify(). - VirtioBlkInit(): jump to Failed label if SetPageSize() fails - VirtioBlkInit(): fixup comment, and add error handling, near SetQueueNum() call - VirtioBlkDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_BLOCK_DEVICE; VirtioBlkDriverBindingSupported() handles it already - VirtioNetGetFeatures(): update stale comment block - VirtioNetGetFeatures(): retrieve MAC address byte for byte (open-coded loop) - VirtioNetDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_NETWORK_CARD; VirtioNetDriverBindingSupported() handles it already - VirtioNetInitRing(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation - VirtioNetInitialize(): fix destination error label for when SetPageSize() fails - VirtioScsi.c: fix comment block of VIRTIO_CFG_WRITE()/VIRTIO_CFG_READ() - VirtioScsiInit(): fix destination error label for when SetPageSize() fails - VirtioScsiInit(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@14966 6f19259b-4bc3-4df7-8a09-765794883524
2013-12-11 17:58:22 +01:00
Portion of Copyright (C) 2013, ARM Ltd.
Copyright (C) 2017, AMD Inc, All rights reserved.<BR>
SPDX-License-Identifier: BSD-2-Clause-Patent
**/
#include <Library/BaseLib.h>
#include <Library/BaseMemoryLib.h>
#include <Library/DebugLib.h>
#include <Library/UefiBootServicesTableLib.h>
#include <Library/VirtioLib.h>
/**
Configure a virtio ring.
This function sets up internal storage (the guest-host communication area)
and lays out several "navigation" (ie. no-ownership) pointers to parts of
that storage.
Relevant sections from the virtio-0.9.5 spec:
- 1.1 Virtqueues,
- 2.3 Virtqueue Configuration.
@param[in] VirtIo The virtio device which will use the ring.
@param[in] The number of descriptors to allocate for the
virtio ring, as requested by the host.
@param[out] Ring The virtio ring to set up.
@return Status codes propagated from
VirtIo->AllocateSharedPages().
@retval EFI_SUCCESS Allocation and setup successful. Ring->Base
(and nothing else) is responsible for
deallocation.
**/
EFI_STATUS
EFIAPI
VirtioRingInit (
IN VIRTIO_DEVICE_PROTOCOL *VirtIo,
IN UINT16 QueueSize,
OUT VRING *Ring
)
{
EFI_STATUS Status;
UINTN RingSize;
volatile UINT8 *RingPagesPtr;
RingSize = ALIGN_VALUE (
sizeof *Ring->Desc * QueueSize +
sizeof *Ring->Avail.Flags +
sizeof *Ring->Avail.Idx +
sizeof *Ring->Avail.Ring * QueueSize +
sizeof *Ring->Avail.UsedEvent,
EFI_PAGE_SIZE);
RingSize += ALIGN_VALUE (
sizeof *Ring->Used.Flags +
sizeof *Ring->Used.Idx +
sizeof *Ring->Used.UsedElem * QueueSize +
sizeof *Ring->Used.AvailEvent,
EFI_PAGE_SIZE);
//
// Allocate a shared ring buffer
//
Ring->NumPages = EFI_SIZE_TO_PAGES (RingSize);
Status = VirtIo->AllocateSharedPages (
VirtIo,
Ring->NumPages,
&Ring->Base
);
if (EFI_ERROR (Status)) {
return Status;
}
SetMem (Ring->Base, RingSize, 0x00);
RingPagesPtr = Ring->Base;
Ring->Desc = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Desc * QueueSize;
Ring->Avail.Flags = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Avail.Flags;
Ring->Avail.Idx = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Avail.Idx;
Ring->Avail.Ring = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Avail.Ring * QueueSize;
Ring->Avail.UsedEvent = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Avail.UsedEvent;
RingPagesPtr = (volatile UINT8 *) Ring->Base +
ALIGN_VALUE (RingPagesPtr - (volatile UINT8 *) Ring->Base,
EFI_PAGE_SIZE);
Ring->Used.Flags = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Used.Flags;
Ring->Used.Idx = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Used.Idx;
Ring->Used.UsedElem = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Used.UsedElem * QueueSize;
Ring->Used.AvailEvent = (volatile VOID *) RingPagesPtr;
RingPagesPtr += sizeof *Ring->Used.AvailEvent;
Ring->QueueSize = QueueSize;
return EFI_SUCCESS;
}
/**
Tear down the internal resources of a configured virtio ring.
The caller is responsible to stop the host from using this ring before
invoking this function: the VSTAT_DRIVER_OK bit must be clear in
VhdrDeviceStatus.
@param[in] VirtIo The virtio device which was using the ring.
@param[out] Ring The virtio ring to clean up.
**/
VOID
EFIAPI
VirtioRingUninit (
IN VIRTIO_DEVICE_PROTOCOL *VirtIo,
IN OUT VRING *Ring
)
{
VirtIo->FreeSharedPages (VirtIo, Ring->NumPages, Ring->Base);
SetMem (Ring, sizeof *Ring, 0x00);
}
/**
Turn off interrupt notifications from the host, and prepare for appending
multiple descriptors to the virtio ring.
The calling driver must be in VSTAT_DRIVER_OK state.
@param[in,out] Ring The virtio ring we intend to append descriptors to.
@param[out] Indices The DESC_INDICES structure to initialize.
**/
VOID
EFIAPI
VirtioPrepare (
IN OUT VRING *Ring,
OUT DESC_INDICES *Indices
)
{
//
// Prepare for virtio-0.9.5, 2.4.2 Receiving Used Buffers From the Device.
// We're going to poll the answer, the host should not send an interrupt.
//
*Ring->Avail.Flags = (UINT16) VRING_AVAIL_F_NO_INTERRUPT;
//
// Prepare for virtio-0.9.5, 2.4.1 Supplying Buffers to the Device.
//
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
// Since we support only one in-flight descriptor chain, we can always build
// that chain starting at entry #0 of the descriptor table.
//
Indices->HeadDescIdx = 0;
Indices->NextDescIdx = Indices->HeadDescIdx;
}
/**
Append a contiguous buffer for transmission / reception via the virtio ring.
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
This function implements the following section from virtio-0.9.5:
- 2.4.1.1 Placing Buffers into the Descriptor Table
Free space is taken as granted, since the individual drivers support only
synchronous requests and host side status is processed in lock-step with
request submission. It is the calling driver's responsibility to verify the
ring size in advance.
The caller is responsible for initializing *Indices with VirtioPrepare()
first.
@param[in,out] Ring The virtio ring to append the buffer to,
as a descriptor.
@param[in] BufferDeviceAddress (Bus master device) start address of the
transmit / receive buffer.
@param[in] BufferSize Number of bytes to transmit or receive.
@param[in] Flags A bitmask of VRING_DESC_F_* flags. The
caller computes this mask dependent on
further buffers to append and transfer
direction. VRING_DESC_F_INDIRECT is
unsupported. The VRING_DESC.Next field is
always set, but the host only interprets
it dependent on VRING_DESC_F_NEXT.
@param[in,out] Indices Indices->HeadDescIdx is not accessed.
On input, Indices->NextDescIdx identifies
the next descriptor to carry the buffer.
On output, Indices->NextDescIdx is
incremented by one, modulo 2^16.
**/
VOID
EFIAPI
VirtioAppendDesc (
IN OUT VRING *Ring,
IN UINT64 BufferDeviceAddress,
IN UINT32 BufferSize,
IN UINT16 Flags,
IN OUT DESC_INDICES *Indices
)
{
volatile VRING_DESC *Desc;
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
Desc = &Ring->Desc[Indices->NextDescIdx++ % Ring->QueueSize];
Desc->Addr = BufferDeviceAddress;
Desc->Len = BufferSize;
Desc->Flags = Flags;
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
Desc->Next = Indices->NextDescIdx % Ring->QueueSize;
}
/**
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
Notify the host about the descriptor chain just built, and wait until the
host processes it.
OvmfPkg: Make the VirtIo devices use the new VIRTIO_DEVICE_PROTOCOL This change replaces the accesses to the PCI bus from the Block, Scsi and Net drivers by the use of the new VIRTIO_DEVICE_PROTOCOL protocol that abstracts the transport layer. It means these drivers can be used on PCI and MMIO transport layer. Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Olivier Martin <olivier.martin@arm.com> v5: - VirtioFlush(): update comment block in VirtioLib.[hc]; error code is propagated from VirtIo->SetQueueNotify(). - VirtioBlkInit(): jump to Failed label if SetPageSize() fails - VirtioBlkInit(): fixup comment, and add error handling, near SetQueueNum() call - VirtioBlkDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_BLOCK_DEVICE; VirtioBlkDriverBindingSupported() handles it already - VirtioNetGetFeatures(): update stale comment block - VirtioNetGetFeatures(): retrieve MAC address byte for byte (open-coded loop) - VirtioNetDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_NETWORK_CARD; VirtioNetDriverBindingSupported() handles it already - VirtioNetInitRing(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation - VirtioNetInitialize(): fix destination error label for when SetPageSize() fails - VirtioScsi.c: fix comment block of VIRTIO_CFG_WRITE()/VIRTIO_CFG_READ() - VirtioScsiInit(): fix destination error label for when SetPageSize() fails - VirtioScsiInit(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@14966 6f19259b-4bc3-4df7-8a09-765794883524
2013-12-11 17:58:22 +01:00
@param[in] VirtIo The target virtio device to notify.
@param[in] VirtQueueId Identifies the queue for the target device.
@param[in,out] Ring The virtio ring with descriptors to submit.
@param[in] Indices Indices->NextDescIdx is not accessed.
Indices->HeadDescIdx identifies the head descriptor
of the descriptor chain.
OvmfPkg: VirtioFlush(): return the number of bytes written by the host VirtioLib provides an API for simple, synchronous (request/response-style) virtio communication. The guest driver builds one descriptor chain, link for link, with VirtioPrepare() and VirtioAppendDesc(), then submits the chain, and awaits the processing, with VirtioFlush(). The descriptor chain is always built at the beginning of the descriptor area, with the head descriptor having descriptor index 0. In order to submit the descriptor chain to the host, the guest always pushes a new "available element" to the Available Ring, in genuine queue-like fashion, with the new element referencing the head descriptor (which always has index 0, see above). In turn, after processing, the host always pushes a new "used element" to the Used Ring, in genuine queue-like fashion, with the new element referencing the head descriptor of the chain that was just processed. The same element also reports the number of bytes that the host wrote, consecutively across the host-writeable buffers that were linked by the descriptors. (See "OvmfPkg/VirtioNetDxe/TechNotes.txt" for a diagram about the descriptor area and the rings.) Because at most one descriptor chain can be in flight with VirtioLib at any time, - the Available Ring and the Used Ring proceed in lock-step, - and the head descriptor that the new "available" and "used" elements can ever reference has index 0. Based on the above, we can modify VirtioFlush() to return the number of bytes written by the host across the descriptor chain. The virtio-block and virtio-scsi drivers don't care (they have other ways to parse the data produced by the host), while the virtio-net driver doesn't use VirtioFlush() at all (it employs VirtioLib only to set up its rings). However, the virtio entropy device, to be covered in the upcoming patches, reports the amount of randomness produced by the host only through this quantity. Cc: Jordan Justen <jordan.l.justen@intel.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-02-19 18:15:07 +01:00
@param[out] UsedLen On success, the total number of bytes, consecutively
across the buffers linked by the descriptor chain,
that the host wrote. May be NULL if the caller
doesn't care, or can compute the same information
from device-specific request structures linked by the
descriptor chain.
OvmfPkg: Make the VirtIo devices use the new VIRTIO_DEVICE_PROTOCOL This change replaces the accesses to the PCI bus from the Block, Scsi and Net drivers by the use of the new VIRTIO_DEVICE_PROTOCOL protocol that abstracts the transport layer. It means these drivers can be used on PCI and MMIO transport layer. Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Olivier Martin <olivier.martin@arm.com> v5: - VirtioFlush(): update comment block in VirtioLib.[hc]; error code is propagated from VirtIo->SetQueueNotify(). - VirtioBlkInit(): jump to Failed label if SetPageSize() fails - VirtioBlkInit(): fixup comment, and add error handling, near SetQueueNum() call - VirtioBlkDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_BLOCK_DEVICE; VirtioBlkDriverBindingSupported() handles it already - VirtioNetGetFeatures(): update stale comment block - VirtioNetGetFeatures(): retrieve MAC address byte for byte (open-coded loop) - VirtioNetDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_NETWORK_CARD; VirtioNetDriverBindingSupported() handles it already - VirtioNetInitRing(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation - VirtioNetInitialize(): fix destination error label for when SetPageSize() fails - VirtioScsi.c: fix comment block of VIRTIO_CFG_WRITE()/VIRTIO_CFG_READ() - VirtioScsiInit(): fix destination error label for when SetPageSize() fails - VirtioScsiInit(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@14966 6f19259b-4bc3-4df7-8a09-765794883524
2013-12-11 17:58:22 +01:00
@return Error code from VirtIo->SetQueueNotify() if it fails.
@retval EFI_SUCCESS Otherwise, the host processed all descriptors.
**/
EFI_STATUS
EFIAPI
VirtioFlush (
OvmfPkg: Make the VirtIo devices use the new VIRTIO_DEVICE_PROTOCOL This change replaces the accesses to the PCI bus from the Block, Scsi and Net drivers by the use of the new VIRTIO_DEVICE_PROTOCOL protocol that abstracts the transport layer. It means these drivers can be used on PCI and MMIO transport layer. Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Olivier Martin <olivier.martin@arm.com> v5: - VirtioFlush(): update comment block in VirtioLib.[hc]; error code is propagated from VirtIo->SetQueueNotify(). - VirtioBlkInit(): jump to Failed label if SetPageSize() fails - VirtioBlkInit(): fixup comment, and add error handling, near SetQueueNum() call - VirtioBlkDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_BLOCK_DEVICE; VirtioBlkDriverBindingSupported() handles it already - VirtioNetGetFeatures(): update stale comment block - VirtioNetGetFeatures(): retrieve MAC address byte for byte (open-coded loop) - VirtioNetDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_NETWORK_CARD; VirtioNetDriverBindingSupported() handles it already - VirtioNetInitRing(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation - VirtioNetInitialize(): fix destination error label for when SetPageSize() fails - VirtioScsi.c: fix comment block of VIRTIO_CFG_WRITE()/VIRTIO_CFG_READ() - VirtioScsiInit(): fix destination error label for when SetPageSize() fails - VirtioScsiInit(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@14966 6f19259b-4bc3-4df7-8a09-765794883524
2013-12-11 17:58:22 +01:00
IN VIRTIO_DEVICE_PROTOCOL *VirtIo,
IN UINT16 VirtQueueId,
IN OUT VRING *Ring,
OvmfPkg: VirtioFlush(): return the number of bytes written by the host VirtioLib provides an API for simple, synchronous (request/response-style) virtio communication. The guest driver builds one descriptor chain, link for link, with VirtioPrepare() and VirtioAppendDesc(), then submits the chain, and awaits the processing, with VirtioFlush(). The descriptor chain is always built at the beginning of the descriptor area, with the head descriptor having descriptor index 0. In order to submit the descriptor chain to the host, the guest always pushes a new "available element" to the Available Ring, in genuine queue-like fashion, with the new element referencing the head descriptor (which always has index 0, see above). In turn, after processing, the host always pushes a new "used element" to the Used Ring, in genuine queue-like fashion, with the new element referencing the head descriptor of the chain that was just processed. The same element also reports the number of bytes that the host wrote, consecutively across the host-writeable buffers that were linked by the descriptors. (See "OvmfPkg/VirtioNetDxe/TechNotes.txt" for a diagram about the descriptor area and the rings.) Because at most one descriptor chain can be in flight with VirtioLib at any time, - the Available Ring and the Used Ring proceed in lock-step, - and the head descriptor that the new "available" and "used" elements can ever reference has index 0. Based on the above, we can modify VirtioFlush() to return the number of bytes written by the host across the descriptor chain. The virtio-block and virtio-scsi drivers don't care (they have other ways to parse the data produced by the host), while the virtio-net driver doesn't use VirtioFlush() at all (it employs VirtioLib only to set up its rings). However, the virtio entropy device, to be covered in the upcoming patches, reports the amount of randomness produced by the host only through this quantity. Cc: Jordan Justen <jordan.l.justen@intel.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-02-19 18:15:07 +01:00
IN DESC_INDICES *Indices,
OUT UINT32 *UsedLen OPTIONAL
)
{
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
UINT16 NextAvailIdx;
OvmfPkg: VirtioFlush(): return the number of bytes written by the host VirtioLib provides an API for simple, synchronous (request/response-style) virtio communication. The guest driver builds one descriptor chain, link for link, with VirtioPrepare() and VirtioAppendDesc(), then submits the chain, and awaits the processing, with VirtioFlush(). The descriptor chain is always built at the beginning of the descriptor area, with the head descriptor having descriptor index 0. In order to submit the descriptor chain to the host, the guest always pushes a new "available element" to the Available Ring, in genuine queue-like fashion, with the new element referencing the head descriptor (which always has index 0, see above). In turn, after processing, the host always pushes a new "used element" to the Used Ring, in genuine queue-like fashion, with the new element referencing the head descriptor of the chain that was just processed. The same element also reports the number of bytes that the host wrote, consecutively across the host-writeable buffers that were linked by the descriptors. (See "OvmfPkg/VirtioNetDxe/TechNotes.txt" for a diagram about the descriptor area and the rings.) Because at most one descriptor chain can be in flight with VirtioLib at any time, - the Available Ring and the Used Ring proceed in lock-step, - and the head descriptor that the new "available" and "used" elements can ever reference has index 0. Based on the above, we can modify VirtioFlush() to return the number of bytes written by the host across the descriptor chain. The virtio-block and virtio-scsi drivers don't care (they have other ways to parse the data produced by the host), while the virtio-net driver doesn't use VirtioFlush() at all (it employs VirtioLib only to set up its rings). However, the virtio entropy device, to be covered in the upcoming patches, reports the amount of randomness produced by the host only through this quantity. Cc: Jordan Justen <jordan.l.justen@intel.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-02-19 18:15:07 +01:00
UINT16 LastUsedIdx;
EFI_STATUS Status;
UINTN PollPeriodUsecs;
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
//
// virtio-0.9.5, 2.4.1.2 Updating the Available Ring
//
// It is not exactly clear from the wording of the virtio-0.9.5
// specification, but each entry in the Available Ring references only the
// head descriptor of any given descriptor chain.
//
NextAvailIdx = *Ring->Avail.Idx;
OvmfPkg: VirtioFlush(): return the number of bytes written by the host VirtioLib provides an API for simple, synchronous (request/response-style) virtio communication. The guest driver builds one descriptor chain, link for link, with VirtioPrepare() and VirtioAppendDesc(), then submits the chain, and awaits the processing, with VirtioFlush(). The descriptor chain is always built at the beginning of the descriptor area, with the head descriptor having descriptor index 0. In order to submit the descriptor chain to the host, the guest always pushes a new "available element" to the Available Ring, in genuine queue-like fashion, with the new element referencing the head descriptor (which always has index 0, see above). In turn, after processing, the host always pushes a new "used element" to the Used Ring, in genuine queue-like fashion, with the new element referencing the head descriptor of the chain that was just processed. The same element also reports the number of bytes that the host wrote, consecutively across the host-writeable buffers that were linked by the descriptors. (See "OvmfPkg/VirtioNetDxe/TechNotes.txt" for a diagram about the descriptor area and the rings.) Because at most one descriptor chain can be in flight with VirtioLib at any time, - the Available Ring and the Used Ring proceed in lock-step, - and the head descriptor that the new "available" and "used" elements can ever reference has index 0. Based on the above, we can modify VirtioFlush() to return the number of bytes written by the host across the descriptor chain. The virtio-block and virtio-scsi drivers don't care (they have other ways to parse the data produced by the host), while the virtio-net driver doesn't use VirtioFlush() at all (it employs VirtioLib only to set up its rings). However, the virtio entropy device, to be covered in the upcoming patches, reports the amount of randomness produced by the host only through this quantity. Cc: Jordan Justen <jordan.l.justen@intel.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-02-19 18:15:07 +01:00
//
// (Due to our lock-step progress, this is where the host will produce the
// used element with the head descriptor's index in it.)
//
LastUsedIdx = NextAvailIdx;
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
Ring->Avail.Ring[NextAvailIdx++ % Ring->QueueSize] =
Indices->HeadDescIdx % Ring->QueueSize;
//
// virtio-0.9.5, 2.4.1.3 Updating the Index Field
//
MemoryFence();
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
*Ring->Avail.Idx = NextAvailIdx;
//
// virtio-0.9.5, 2.4.1.4 Notifying the Device -- gratuitous notifications are
// OK.
//
MemoryFence();
OvmfPkg: Make the VirtIo devices use the new VIRTIO_DEVICE_PROTOCOL This change replaces the accesses to the PCI bus from the Block, Scsi and Net drivers by the use of the new VIRTIO_DEVICE_PROTOCOL protocol that abstracts the transport layer. It means these drivers can be used on PCI and MMIO transport layer. Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Olivier Martin <olivier.martin@arm.com> v5: - VirtioFlush(): update comment block in VirtioLib.[hc]; error code is propagated from VirtIo->SetQueueNotify(). - VirtioBlkInit(): jump to Failed label if SetPageSize() fails - VirtioBlkInit(): fixup comment, and add error handling, near SetQueueNum() call - VirtioBlkDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_BLOCK_DEVICE; VirtioBlkDriverBindingSupported() handles it already - VirtioNetGetFeatures(): update stale comment block - VirtioNetGetFeatures(): retrieve MAC address byte for byte (open-coded loop) - VirtioNetDriverBindingStart(): remove redundant (always false) check for a subsystem device ID different from VIRTIO_SUBSYSTEM_NETWORK_CARD; VirtioNetDriverBindingSupported() handles it already - VirtioNetInitRing(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation - VirtioNetInitialize(): fix destination error label for when SetPageSize() fails - VirtioScsi.c: fix comment block of VIRTIO_CFG_WRITE()/VIRTIO_CFG_READ() - VirtioScsiInit(): fix destination error label for when SetPageSize() fails - VirtioScsiInit(): call SetQueueNum() and SetQueueAlign() for proper MMIO operation Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@14966 6f19259b-4bc3-4df7-8a09-765794883524
2013-12-11 17:58:22 +01:00
Status = VirtIo->SetQueueNotify (VirtIo, VirtQueueId);
if (EFI_ERROR (Status)) {
return Status;
}
//
// virtio-0.9.5, 2.4.2 Receiving Used Buffers From the Device
// Wait until the host processes and acknowledges our descriptor chain. The
// condition we use for polling is greatly simplified and relies on the
// synchronous, lock-step progress.
//
// Keep slowing down until we reach a poll period of slightly above 1 ms.
//
PollPeriodUsecs = 1;
MemoryFence();
OvmfPkg: VirtioLib: populate the Available Ring correctly The descriptor table (also known as "queue") consists of descriptors. (The corresponding type in the code is VRING_DESC.) An individual descriptor describes a contiguous buffer, to be transferred uni-directionally between host and guest. Several descriptors in the descriptor table can be linked into a descriptor chain, specifying a bi-directional scatter-gather transfer between host and guest. Such a descriptor chain is also known as "virtio request". (The descriptor table can host sereval descriptor chains (in-flight virtio requests) in parallel, but the OVMF driver supports at most one chain, at any point in time.) The first descriptor in any descriptor chain is called "head descriptor". In order to submit a number of parallel requests (= a set of independent descriptor chains) from the guest to the host, the guest must put *only* the head descriptor of each separate chain onto the Available Ring. VirtioLib currently places the head of its one descriptor chain onto the Available Ring repeatedly, once for each single (head *or* dependent) descriptor in said descriptor chain. If the descriptor chain comprises N descriptors, this error amounts to submitting the same entire chain N times in parallel. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Ptr to head --/ Desc#1 (next in same chain) ... / ... Ptr to head / Desc#(N-1) (last in same chain) Anatomy of a single virtio-blk READ request (a descriptor chain with three descriptors): virtio-blk request header, prepared by guest: VirtioAppendDesc PhysAddr=3FBC6050 Size=16 Flags=1 Head=1232 Next=1232 payload to be filled in by host: VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=1232 Next=1233 host status, to be filled in by host: VirtioAppendDesc PhysAddr=3FBC604F Size=1 Flags=2 Head=1232 Next=1234 Processing on the host side -- the descriptor chain is processed three times in parallel (its head is available to virtqueue_pop() thrice); the same chain is submitted/collected separately to/from AIO three times: virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtqueue_pop vq VQ#0 elem EL#1 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#1 virtqueue_pop vq VQ#0 elem EL#2 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#2 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 virtio_blk_rw_complete req REQ#1 ret 0 virtio_blk_req_complete req REQ#1 status 0 virtio_blk_rw_complete req REQ#2 ret 0 virtio_blk_req_complete req REQ#2 status 0 On my Thinkpad T510 laptop with RHEL-6 as host, this probably leads to simultaneous DMA transfers targeting the same RAM area. Even though the source of each transfer is identical, the data is corrupted in the destination buffer -- the CRC32 calculated over the buffer varies, even though the origin of the transfers is the same, never rewritten LBA. SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=BF68A44D The problem is invisible on my HP Z400 workstation. Fix the request submission by: - building the only one descriptor chain supported by VirtioLib always at the beginning of the descriptor table, - ensuring the head descriptor of this chain is put on the Available Ring only once, - requesting the virtio spec's language to be cleaned up <http://lists.linuxfoundation.org/pipermail/virtualization/2013-April/024032.html>. Available Ring Descriptor table Ptr to head ----> Desc#0 (head of chain) Desc#1 (next in same chain) ... Desc#(N-1) (last in same chain) VirtioAppendDesc PhysAddr=3FBC6040 Size=16 Flags=1 Head=0 Next=0 VirtioAppendDesc PhysAddr=3B934C00 Size=32768 Flags=3 Head=0 Next=1 VirtioAppendDesc PhysAddr=3FBC603F Size=1 Flags=2 Head=0 Next=2 virtio_queue_notify vdev VDEV vq VQ#0 virtqueue_pop vq VQ#0 elem EL#0 in_num 2 out_num 1 bdrv_aio_readv bs BDRV sector_num 585792 nb_sectors 64 opaque REQ#0 virtio_blk_rw_complete req REQ#0 ret 0 virtio_blk_req_complete req REQ#0 status 0 SynchronousRequest Lba=585792 BufSiz=32768 ReqIsWrite=0 Crc32=1EEB2B07 (The Crc32 was double-checked with edk2's and Linux's guest IDE driver.) Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14356 6f19259b-4bc3-4df7-8a09-765794883524
2013-05-14 17:57:55 +02:00
while (*Ring->Used.Idx != NextAvailIdx) {
gBS->Stall (PollPeriodUsecs); // calls AcpiTimerLib::MicroSecondDelay
if (PollPeriodUsecs < 1024) {
PollPeriodUsecs *= 2;
}
MemoryFence();
}
MemoryFence();
OvmfPkg: VirtioFlush(): return the number of bytes written by the host VirtioLib provides an API for simple, synchronous (request/response-style) virtio communication. The guest driver builds one descriptor chain, link for link, with VirtioPrepare() and VirtioAppendDesc(), then submits the chain, and awaits the processing, with VirtioFlush(). The descriptor chain is always built at the beginning of the descriptor area, with the head descriptor having descriptor index 0. In order to submit the descriptor chain to the host, the guest always pushes a new "available element" to the Available Ring, in genuine queue-like fashion, with the new element referencing the head descriptor (which always has index 0, see above). In turn, after processing, the host always pushes a new "used element" to the Used Ring, in genuine queue-like fashion, with the new element referencing the head descriptor of the chain that was just processed. The same element also reports the number of bytes that the host wrote, consecutively across the host-writeable buffers that were linked by the descriptors. (See "OvmfPkg/VirtioNetDxe/TechNotes.txt" for a diagram about the descriptor area and the rings.) Because at most one descriptor chain can be in flight with VirtioLib at any time, - the Available Ring and the Used Ring proceed in lock-step, - and the head descriptor that the new "available" and "used" elements can ever reference has index 0. Based on the above, we can modify VirtioFlush() to return the number of bytes written by the host across the descriptor chain. The virtio-block and virtio-scsi drivers don't care (they have other ways to parse the data produced by the host), while the virtio-net driver doesn't use VirtioFlush() at all (it employs VirtioLib only to set up its rings). However, the virtio entropy device, to be covered in the upcoming patches, reports the amount of randomness produced by the host only through this quantity. Cc: Jordan Justen <jordan.l.justen@intel.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-02-19 18:15:07 +01:00
if (UsedLen != NULL) {
volatile CONST VRING_USED_ELEM *UsedElem;
UsedElem = &Ring->Used.UsedElem[LastUsedIdx % Ring->QueueSize];
ASSERT (UsedElem->Id == Indices->HeadDescIdx);
*UsedLen = UsedElem->Len;
}
return EFI_SUCCESS;
}
/**
Report the feature bits to the VirtIo 1.0 device that the VirtIo 1.0 driver
understands.
In VirtIo 1.0, a device can reject a self-inconsistent feature bitmap through
the new VSTAT_FEATURES_OK status bit. (For example if the driver requests a
higher level feature but clears a prerequisite feature.) This function is a
small wrapper around VIRTIO_DEVICE_PROTOCOL.SetGuestFeatures() that also
verifies if the VirtIo 1.0 device accepts the feature bitmap.
@param[in] VirtIo Report feature bits to this device.
@param[in] Features The set of feature bits that the driver wishes
to report. The caller is responsible to perform
any masking before calling this function; the
value is directly written with
VIRTIO_DEVICE_PROTOCOL.SetGuestFeatures().
@param[in,out] DeviceStatus On input, the status byte most recently written
to the device's status register. On output (even
on error), DeviceStatus will be updated so that
it is suitable for further status bit
manipulation and writing to the device's status
register.
@retval EFI_SUCCESS The device accepted the configuration in Features.
@return EFI_UNSUPPORTED The device rejected the configuration in Features.
@retval EFI_UNSUPPORTED VirtIo->Revision is smaller than 1.0.0.
@return Error codes from the SetGuestFeatures(),
SetDeviceStatus(), GetDeviceStatus() member
functions.
**/
EFI_STATUS
EFIAPI
Virtio10WriteFeatures (
IN VIRTIO_DEVICE_PROTOCOL *VirtIo,
IN UINT64 Features,
IN OUT UINT8 *DeviceStatus
)
{
EFI_STATUS Status;
if (VirtIo->Revision < VIRTIO_SPEC_REVISION (1, 0, 0)) {
return EFI_UNSUPPORTED;
}
Status = VirtIo->SetGuestFeatures (VirtIo, Features);
if (EFI_ERROR (Status)) {
return Status;
}
*DeviceStatus |= VSTAT_FEATURES_OK;
Status = VirtIo->SetDeviceStatus (VirtIo, *DeviceStatus);
if (EFI_ERROR (Status)) {
return Status;
}
Status = VirtIo->GetDeviceStatus (VirtIo, DeviceStatus);
if (EFI_ERROR (Status)) {
return Status;
}
if ((*DeviceStatus & VSTAT_FEATURES_OK) == 0) {
Status = EFI_UNSUPPORTED;
}
return Status;
}
/**
Provides the virtio device address required to access system memory from a
DMA bus master.
The interface follows the same usage pattern as defined in UEFI spec 2.6
(Section 13.2 PCI Root Bridge I/O Protocol)
The VirtioMapAllBytesInSharedBuffer() is similar to VIRTIO_MAP_SHARED
with exception that NumberOfBytes is IN-only parameter. The function
maps all the bytes specified in NumberOfBytes param in one consecutive
range.
@param[in] VirtIo The virtio device for which the mapping is
requested.
@param[in] Operation Indicates if the bus master is going to
read or write to system memory.
@param[in] HostAddress The system memory address to map to shared
buffer address.
@param[in] NumberOfBytes Number of bytes to map.
@param[out] DeviceAddress The resulting shared map address for the
bus master to access the hosts HostAddress.
@param[out] Mapping A resulting token to pass to
VIRTIO_UNMAP_SHARED.
@retval EFI_SUCCESS The NumberOfBytes is successfully mapped.
@retval EFI_UNSUPPORTED The HostAddress cannot be mapped as a
common buffer.
@retval EFI_INVALID_PARAMETER One or more parameters are invalid.
@retval EFI_OUT_OF_RESOURCES The request could not be completed due to
a lack of resources. This includes the case
when NumberOfBytes bytes cannot be mapped
in one consecutive range.
@retval EFI_DEVICE_ERROR The system hardware could not map the
requested address.
**/
EFI_STATUS
EFIAPI
VirtioMapAllBytesInSharedBuffer (
IN VIRTIO_DEVICE_PROTOCOL *VirtIo,
IN VIRTIO_MAP_OPERATION Operation,
IN VOID *HostAddress,
IN UINTN NumberOfBytes,
OUT EFI_PHYSICAL_ADDRESS *DeviceAddress,
OUT VOID **Mapping
)
{
EFI_STATUS Status;
VOID *MapInfo;
UINTN Size;
EFI_PHYSICAL_ADDRESS PhysicalAddress;
Size = NumberOfBytes;
Status = VirtIo->MapSharedBuffer (
VirtIo,
Operation,
HostAddress,
&Size,
&PhysicalAddress,
&MapInfo
);
if (EFI_ERROR (Status)) {
return Status;
}
if (Size < NumberOfBytes) {
goto Failed;
}
*Mapping = MapInfo;
*DeviceAddress = PhysicalAddress;
return EFI_SUCCESS;
Failed:
VirtIo->UnmapSharedBuffer (VirtIo, MapInfo);
return EFI_OUT_OF_RESOURCES;
}
/**
Map the ring buffer so that it can be accessed equally by both guest
and hypervisor.
@param[in] VirtIo The virtio device instance.
@param[in] Ring The virtio ring to map.
@param[out] RingBaseShift A resulting translation offset, to be
passed to VirtIo->SetQueueAddress().
@param[out] Mapping A resulting token to pass to
VirtIo->UnmapSharedBuffer().
@return Status code from VirtIo->MapSharedBuffer()
**/
EFI_STATUS
EFIAPI
VirtioRingMap (
IN VIRTIO_DEVICE_PROTOCOL *VirtIo,
IN VRING *Ring,
OUT UINT64 *RingBaseShift,
OUT VOID **Mapping
)
{
EFI_STATUS Status;
EFI_PHYSICAL_ADDRESS DeviceAddress;
Status = VirtioMapAllBytesInSharedBuffer (
VirtIo,
VirtioOperationBusMasterCommonBuffer,
Ring->Base,
EFI_PAGES_TO_SIZE (Ring->NumPages),
&DeviceAddress,
Mapping
);
if (EFI_ERROR (Status)) {
return Status;
}
*RingBaseShift = DeviceAddress - (UINT64)(UINTN)Ring->Base;
return EFI_SUCCESS;
}