Several fixes in this patch:
* Don't ignore function level and per-vector masking. We're not
supposed to signal when masked and not doing so will improve
performance a bit (in addition to behaving correctly).
* Implement the missing PBA array. 'lspci -vv' will now show the correct
output:
Capabilities: [40] MSI-X: Enable+ Count=5 Masked-
Vector table: BAR=1 offset=00000000
PBA: BAR=3 offset=00000000
* Checking whether MSI-X is enabled or not is done by probing the
corresponding bit within the PCI header instead of trying to track
the status ourselves.
* Fallback to INTx if MSI-X isn't enabled.
* Use correct max size for MSI-X table.
Signed-off-by: Sasha Levin <levinsasha928@xxxxxxxxx>
---
tools/kvm/include/kvm/pci.h | 4 +-
tools/kvm/include/kvm/virtio-pci.h | 4 +-
tools/kvm/virtio/pci.c | 89 ++++++++++++++++++++++++++++--------
3 files changed, 74 insertions(+), 23 deletions(-)
diff --git a/tools/kvm/include/kvm/pci.h b/tools/kvm/include/kvm/pci.h
index 2ab5291..e74d3ec 100644
--- a/tools/kvm/include/kvm/pci.h
+++ b/tools/kvm/include/kvm/pci.h
@@ -34,9 +34,9 @@ struct msix_table {
struct msix_cap {
u8 cap;
u8 next;
- u16 table_size;
+ u16 ctrl;
u32 table_offset;
- struct msix_table table[3];
+ u32 pba_offset;
};
struct pci_device_header {
diff --git a/tools/kvm/include/kvm/virtio-pci.h b/tools/kvm/include/kvm/virtio-pci.h
index 4524a7f..0c2a035 100644
--- a/tools/kvm/include/kvm/virtio-pci.h
+++ b/tools/kvm/include/kvm/virtio-pci.h
@@ -37,7 +37,9 @@ struct virtio_pci {
u32 vq_vector[VIRTIO_PCI_MAX_VQ];
u32 gsis[VIRTIO_PCI_MAX_VQ];
u32 msix_io_block;
- int msix_enabled;
+ u32 msix_pba_block;
+ u64 msix_pba;
+ struct msix_table msix_table[VIRTIO_PCI_MAX_VQ + 1];
/* virtio queue */
u16 queue_selector;
diff --git a/tools/kvm/virtio/pci.c b/tools/kvm/virtio/pci.c
index 6d086fa..7181123 100644
--- a/tools/kvm/virtio/pci.c
+++ b/tools/kvm/virtio/pci.c
@@ -9,11 +9,17 @@
#include <linux/virtio_pci.h>
#include <string.h>
+static inline bool virtio_pci__msix_enabled(struct virtio_pci *vpci)
+{
+ return vpci->pci_hdr.msix.ctrl & PCI_MSIX_FLAGS_ENABLE;
+}
+
static bool virtio_pci__specific_io_in(struct kvm *kvm, struct virtio_pci *vpci, u16 port,
void *data, int size, int offset)
{
u32 config_offset;
- int type = virtio__get_dev_specific_field(offset - 20, vpci->msix_enabled,
+ int type = virtio__get_dev_specific_field(offset - 20,
+ virtio_pci__msix_enabled(vpci),
0, &config_offset);
if (type == VIRTIO_PCI_O_MSIX) {
switch (offset) {
@@ -81,7 +87,7 @@ static bool virtio_pci__specific_io_out(struct kvm *kvm, struct virtio_pci *vpci
void *data, int size, int offset)
{
u32 config_offset, gsi, vec;
- int type = virtio__get_dev_specific_field(offset - 20, vpci->msix_enabled,
+ int type = virtio__get_dev_specific_field(offset - 20, virtio_pci__msix_enabled(vpci),
0, &config_offset);
if (type == VIRTIO_PCI_O_MSIX) {
switch (offset) {
@@ -89,9 +95,9 @@ static bool virtio_pci__specific_io_out(struct kvm *kvm, struct virtio_pci *vpci
vec = vpci->config_vector = ioport__read16(data);
gsi = irq__add_msix_route(kvm,
- vpci->pci_hdr.msix.table[vec].low,
- vpci->pci_hdr.msix.table[vec].high,
- vpci->pci_hdr.msix.table[vec].data);
+ vpci->msix_table[vec].low,
+ vpci->msix_table[vec].high,
+ vpci->msix_table[vec].data);
vpci->config_gsi = gsi;
break;
@@ -99,9 +105,9 @@ static bool virtio_pci__specific_io_out(struct kvm *kvm, struct virtio_pci *vpci
vec = vpci->vq_vector[vpci->queue_selector] = ioport__read16(data);
gsi = irq__add_msix_route(kvm,
- vpci->pci_hdr.msix.table[vec].low,
- vpci->pci_hdr.msix.table[vec].high,
- vpci->pci_hdr.msix.table[vec].data);
+ vpci->msix_table[vec].low,
+ vpci->msix_table[vec].high,
+ vpci->msix_table[vec].data);
vpci->gsis[vpci->queue_selector] = gsi;
break;
}
@@ -159,28 +165,64 @@ static struct ioport_operations virtio_pci__io_ops = {
.io_out = virtio_pci__io_out,
};
-static void callback_mmio(u64 addr, u8 *data, u32 len, u8 is_write, void *ptr)
+static void callback_mmio_table(u64 addr, u8 *data, u32 len, u8 is_write, void *ptr)
{
struct virtio_pci *vpci = ptr;
- void *table = &vpci->pci_hdr.msix.table;
+ void *table = &vpci->msix_table;
- vpci->msix_enabled = 1;
if (is_write)
memcpy(table + addr - vpci->msix_io_block, data, len);
else
memcpy(data, table + addr - vpci->msix_io_block, len);
}
+static void callback_mmio_pba(u64 addr, u8 *data, u32 len, u8 is_write, void *ptr)
+{
+ struct virtio_pci *vpci = ptr;
+ void *pba = &vpci->msix_pba;
+
+ if (is_write)
+ memcpy(pba + addr - vpci->msix_pba_block, data, len);
+ else
+ memcpy(data, pba + addr - vpci->msix_pba_block, len);
+}
+
int virtio_pci__signal_vq(struct kvm *kvm, struct virtio_pci *vpci, u32 vq)
{
- kvm__irq_line(kvm, vpci->gsis[vq], VIRTIO_IRQ_HIGH);
+ int tbl = vpci->vq_vector[vq];
+ if (virtio_pci__msix_enabled(vpci)) {
+ if (vpci->pci_hdr.msix.ctrl & PCI_MSIX_FLAGS_MASKALL ||
+ vpci->msix_table[tbl].ctrl & PCI_MSIX_ENTRY_CTRL_MASKBIT) {
+
+ vpci->msix_pba |= 1 << tbl;
+ return 0;
+ }
+
+ kvm__irq_trigger(kvm, vpci->gsis[vq]);
+ } else {
+ kvm__irq_trigger(kvm, vpci->pci_hdr.irq_line);
+ }
return 0;
}
int virtio_pci__signal_config(struct kvm *kvm, struct virtio_pci *vpci)
{
- kvm__irq_line(kvm, vpci->config_gsi, VIRTIO_IRQ_HIGH);
+ int tbl = vpci->config_vector;
+
+ if (virtio_pci__msix_enabled(vpci)) {
+ if (vpci->pci_hdr.msix.ctrl & PCI_MSIX_FLAGS_MASKALL ||
+ vpci->msix_table[tbl].ctrl & PCI_MSIX_ENTRY_CTRL_MASKBIT) {
+
+ vpci->msix_pba |= 1 << tbl;
+ return 0;
+ }
+
+ kvm__irq_trigger(kvm, vpci->config_gsi);
+ } else {
+ vpci->isr = VIRTIO_PCI_ISR_CONFIG;
+ kvm__irq_trigger(kvm, vpci->pci_hdr.irq_line);
+ }
return 0;
}
@@ -192,9 +234,11 @@ int virtio_pci__init(struct kvm *kvm, struct virtio_pci *vpci, void *dev,
vpci->dev = dev;
vpci->msix_io_block = pci_get_io_space_block();
+ vpci->msix_pba_block = pci_get_io_space_block();
vpci->base_addr = ioport__register(IOPORT_EMPTY, &virtio_pci__io_ops, IOPORT_SIZE, vpci);
- kvm__register_mmio(kvm, vpci->msix_io_block, 0x100, callback_mmio, vpci);
+ kvm__register_mmio(kvm, vpci->msix_io_block, 0x100, callback_mmio_table, vpci);
+ kvm__register_mmio(kvm, vpci->msix_pba_block, 0x100, callback_mmio_pba, vpci);
vpci->pci_hdr = (struct pci_device_header) {
.vendor_id = PCI_VENDOR_ID_REDHAT_QUMRANET,
@@ -205,19 +249,24 @@ int virtio_pci__init(struct kvm *kvm, struct virtio_pci *vpci, void *dev,
.subsys_vendor_id = PCI_SUBSYSTEM_VENDOR_ID_REDHAT_QUMRANET,
.subsys_id = subsys_id,
.bar[0] = vpci->base_addr | PCI_BASE_ADDRESS_SPACE_IO,
- .bar[1] = vpci->msix_io_block |
- PCI_BASE_ADDRESS_SPACE_MEMORY |
- PCI_BASE_ADDRESS_MEM_TYPE_64,
- /* bar[2] is the continuation of bar[1] for 64bit addressing */
- .bar[2] = 0,
+ .bar[1] = vpci->msix_io_block | PCI_BASE_ADDRESS_SPACE_MEMORY
+ | PCI_BASE_ADDRESS_MEM_TYPE_64,
+ .bar[3] = vpci->msix_pba_block | PCI_BASE_ADDRESS_SPACE_MEMORY
+ | PCI_BASE_ADDRESS_MEM_TYPE_64,
.status = PCI_STATUS_CAP_LIST,
.capabilities = (void *)&vpci->pci_hdr.msix - (void *)&vpci->pci_hdr,
};
vpci->pci_hdr.msix.cap = PCI_CAP_ID_MSIX;
vpci->pci_hdr.msix.next = 0;
- vpci->pci_hdr.msix.table_size = (VIRTIO_PCI_MAX_VQ + 1) | PCI_MSIX_FLAGS_ENABLE;
+ vpci->pci_hdr.msix.ctrl = (VIRTIO_PCI_MAX_VQ + 1);
+
+ /*
+ * Both table and PBA could be mapped on the same BAR, but for now
+ * we're not in short of BARs
+ */
vpci->pci_hdr.msix.table_offset = 1; /* Use BAR 1 */
+ vpci->pci_hdr.msix.pba_offset = 3; /* Use BAR 3 */
vpci->config_vector = 0;
if (irq__register_device(VIRTIO_ID_RNG, &ndev, &pin, &line) < 0)
--
1.7.6
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
