Skip to contentSkip to conversation

msix: couple fixes

PATCH In review Alex Williamson 4 participants 15 messages Oct 22, 2010 – Nov 1, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover msix: couple fixes

I've been porting the qemu vfio driver to use the common msix infrastructure and hit a couple issues. The first patch addresses the problem that a device may already have MSI-X capability setup and non-adjustable BARs, so msix_init should be able to use the existing capability instead of adding it's own.

I'm surprised we need the second patch, as I'm not sure how things work today without it. We allocate msix_irq_entries within a KVM_CAP_IRQCHIP #ifdef. This should be pulled in from kvm.h, if we have CONFIG_KVM, but that's pulled in from config-target.h, which we don't include. We can add config.h to msix.c, but that only works if the msix object is in Makefile.target instead of Makefile.objs. Feel free to fix this differently, but I think something is wrong there. Thanks,

Alex

---

Alex Williamson (2):
msix: Pull in config.h for CONFIG_KVM
msix: Allow msix_init on a device with existing MSI-X capability

 Makefile.objs   |    2 --
 Makefile.target |    1 +
 hw/msix.c       |   68 ++++++++++++++++++++++++++++++++-----------------------
 3 files changed, 40 insertions(+), 31 deletions(-)
#1 lore raw

PATCH 1/2 msix: Allow msix_init on a device with existing MSI-X capability

To enable common msix support to be used with pass through devices, don't attempt to change the BAR if the device already has an MSI-X capability. This also means we want to pay closer attention to the size when we map the msix table page, as it isn't necessarily covering the entire end of the BAR.

---

 hw/msix.c |   67 +++++++++++++++++++++++++++++++++++--------------------------
 1 files changed, 38 insertions(+), 29 deletions(-)
hw/msix.c +38 -29
index 43efbd2..4122395 100644
@@ -167,35 +167,43 @@ static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,
 {
     int config_offset;
     uint8_t *config;
-    uint32_t new_size;
 
-    if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
-        return -EINVAL;
-    if (bar_size > 0x80000000)
-        return -ENOSPC;
-
-    /* Add space for MSI-X structures */
-    if (!bar_size) {
-        new_size = MSIX_PAGE_SIZE;
-    } else if (bar_size < MSIX_PAGE_SIZE) {
-        bar_size = MSIX_PAGE_SIZE;
-        new_size = MSIX_PAGE_SIZE * 2;
-    } else {
-        new_size = bar_size * 2;
-    }
-
-    pdev->msix_bar_size = new_size;
-    config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
-    if (config_offset < 0)
-        return config_offset;
-    config = pdev->config + config_offset;
-
-    pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
-    /* Table on top of BAR */
-    pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
-    /* Pending bits on top of that */
-    pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
-                 bar_nr);
+    pdev->msix_bar_size = bar_size;
+
+    config_offset = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
+
+    if (!config_offset) {
+        uint32_t new_size;
+
+        if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
+            return -EINVAL;
+        if (bar_size > 0x80000000)
+            return -ENOSPC;
+
+        /* Add space for MSI-X structures */
+        if (!bar_size) {
+            new_size = MSIX_PAGE_SIZE;
+        } else if (bar_size < MSIX_PAGE_SIZE) {
+            bar_size = MSIX_PAGE_SIZE;
+            new_size = MSIX_PAGE_SIZE * 2;
+        } else {
+            new_size = bar_size * 2;
+        }
+
+        pdev->msix_bar_size = new_size;
+        config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
+                                           MSIX_CAP_LENGTH);
+        if (config_offset < 0)
+            return config_offset;
+        config = pdev->config + config_offset;
+
+        pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
+        /* Table on top of BAR */
+        pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
+        /* Pending bits on top of that */
+        pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
+                     bar_nr);
+    }
     pdev->msix_cap = config_offset;
     /* Make flags bit writeable. */
     pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
@@ -337,7 +345,8 @@ void msix_mmio_map(PCIDevice *d, int region_num,
         return;
     if (size <= offset)
         return;
-    cpu_register_physical_memory(addr + offset, size - offset,
+    cpu_register_physical_memory(addr + offset,
+                                 MIN(size - offset, MSIX_PAGE_SIZE),
                                  d->msix_mmio_index);
 }
 
#2 lore raw

Re: [PATCH 1/2] msix: Allow msix_init on a device with existing MSI-X capability

Quoted 89 lines · Alex Williamson

To enable common msix support to be used with pass through devices, don't attempt to change the BAR if the device already has an MSI-X capability. This also means we want to pay closer attention to the size when we map the msix table page, as it isn't necessarily covering the entire end of the BAR.

---

 hw/msix.c |   67 +++++++++++++++++++++++++++++++++++--------------------------
 1 files changed, 38 insertions(+), 29 deletions(-)
hw/msix.c +36 -28
index 43efbd2..4122395 100644
@@ -167,35 +167,43 @@ static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,
 {
     int config_offset;
     uint8_t *config;
-    uint32_t new_size;
 
-    if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
-        return -EINVAL;
-    if (bar_size > 0x80000000)
-        return -ENOSPC;
-
-    /* Add space for MSI-X structures */
-    if (!bar_size) {
-        new_size = MSIX_PAGE_SIZE;
-    } else if (bar_size < MSIX_PAGE_SIZE) {
-        bar_size = MSIX_PAGE_SIZE;
-        new_size = MSIX_PAGE_SIZE * 2;
-    } else {
-        new_size = bar_size * 2;
-    }
-
-    pdev->msix_bar_size = new_size;
-    config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
-    if (config_offset < 0)
-        return config_offset;
-    config = pdev->config + config_offset;
-
-    pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
-    /* Table on top of BAR */
-    pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
-    /* Pending bits on top of that */
-    pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
-                 bar_nr);
+    pdev->msix_bar_size = bar_size;
+
+    config_offset = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
+
+    if (!config_offset) {
+        uint32_t new_size;
+
+        if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
+            return -EINVAL;
+        if (bar_size > 0x80000000)
+            return -ENOSPC;
+
+        /* Add space for MSI-X structures */
+        if (!bar_size) {
+            new_size = MSIX_PAGE_SIZE;
+        } else if (bar_size < MSIX_PAGE_SIZE) {
+            bar_size = MSIX_PAGE_SIZE;
+            new_size = MSIX_PAGE_SIZE * 2;
+        } else {
+            new_size = bar_size * 2;
+        }
+
+        pdev->msix_bar_size = new_size;
+        config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
+                                           MSIX_CAP_LENGTH);
+        if (config_offset < 0)
+            return config_offset;
+        config = pdev->config + config_offset;
+
+        pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
+        /* Table on top of BAR */
+        pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
+        /* Pending bits on top of that */
+        pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
+                     bar_nr);
+    }
     pdev->msix_cap = config_offset;
     /* Make flags bit writeable. */
     pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |

hw/msix.c void msix_mmio_map(PCIDevice *d, int region_num, · line 349

Quoted 6 lines
hw/msix.c +2 -1
@@ -337,7 +345,8 @@ void msix_mmio_map(PCIDevice *d, int region_num,
         return;
     if (size <= offset)
         return;
-    cpu_register_physical_memory(addr + offset, size - offset,
+    cpu_register_physical_memory(addr + offset,
+                                 MIN(size - offset, MSIX_PAGE_SIZE),

This is wrong I think, the table might not fit in a single page. You would need to read table size out of from device config.

Quoted 3 lines
hw/msix.c +0 -0
                                  d->msix_mmio_index);
 }
 
#3 lore raw

Re: [PATCH 1/2] msix: Allow msix_init on a device with existing MSI-X capability

Quoted 91 lines · Michael S. Tsirkin
Quoted 90 lines · Alex Williamson

To enable common msix support to be used with pass through devices, don't attempt to change the BAR if the device already has an MSI-X capability. This also means we want to pay closer attention to the size when we map the msix table page, as it isn't necessarily covering the entire end of the BAR.

---

 hw/msix.c |   67 +++++++++++++++++++++++++++++++++++--------------------------
 1 files changed, 38 insertions(+), 29 deletions(-)
hw/msix.c +36 -28
index 43efbd2..4122395 100644
@@ -167,35 +167,43 @@ static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,
 {
     int config_offset;
     uint8_t *config;
-    uint32_t new_size;
 
-    if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
-        return -EINVAL;
-    if (bar_size > 0x80000000)
-        return -ENOSPC;
-
-    /* Add space for MSI-X structures */
-    if (!bar_size) {
-        new_size = MSIX_PAGE_SIZE;
-    } else if (bar_size < MSIX_PAGE_SIZE) {
-        bar_size = MSIX_PAGE_SIZE;
-        new_size = MSIX_PAGE_SIZE * 2;
-    } else {
-        new_size = bar_size * 2;
-    }
-
-    pdev->msix_bar_size = new_size;
-    config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
-    if (config_offset < 0)
-        return config_offset;
-    config = pdev->config + config_offset;
-
-    pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
-    /* Table on top of BAR */
-    pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
-    /* Pending bits on top of that */
-    pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
-                 bar_nr);
+    pdev->msix_bar_size = bar_size;
+
+    config_offset = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
+
+    if (!config_offset) {
+        uint32_t new_size;
+
+        if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
+            return -EINVAL;
+        if (bar_size > 0x80000000)
+            return -ENOSPC;
+
+        /* Add space for MSI-X structures */
+        if (!bar_size) {
+            new_size = MSIX_PAGE_SIZE;
+        } else if (bar_size < MSIX_PAGE_SIZE) {
+            bar_size = MSIX_PAGE_SIZE;
+            new_size = MSIX_PAGE_SIZE * 2;
+        } else {
+            new_size = bar_size * 2;
+        }
+
+        pdev->msix_bar_size = new_size;
+        config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
+                                           MSIX_CAP_LENGTH);
+        if (config_offset < 0)
+            return config_offset;
+        config = pdev->config + config_offset;
+
+        pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
+        /* Table on top of BAR */
+        pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
+        /* Pending bits on top of that */
+        pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
+                     bar_nr);
+    }
     pdev->msix_cap = config_offset;
     /* Make flags bit writeable. */
     pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
@@ -337,7 +345,8 @@ void msix_mmio_map(PCIDevice *d, int region_num,
         return;
Quoted 8 lines

hw/msix.c void msix_mmio_map(PCIDevice *d, int region_num, · line 349

Quoted 5 lines
hw/msix.c +2 -1
     if (size <= offset)
         return;
-    cpu_register_physical_memory(addr + offset, size - offset,
+    cpu_register_physical_memory(addr + offset,
+                                 MIN(size - offset, MSIX_PAGE_SIZE),

This is wrong I think, the table might not fit in a single page. You would need to read table size out of from device config.

That's true, but I was hoping to save that for later since we don't seem to be running into that problem yet. Current device assignment code assumes a single page, and I haven't heard of anyone with a vector table that exceeds that yet. Thanks,

Alex

#4 lore raw

Re: [PATCH 1/2] msix: Allow msix_init on a device with existing MSI-X capability

Quoted 98 lines · Alex Williamson
Quoted 97 lines · Michael S. Tsirkin
Quoted 95 lines · Alex Williamson

To enable common msix support to be used with pass through devices, don't attempt to change the BAR if the device already has an MSI-X capability. This also means we want to pay closer attention to the size when we map the msix table page, as it isn't necessarily covering the entire end of the BAR.

---

  hw/msix.c |   67 +++++++++++++++++++++++++++++++++++--------------------------
  1 files changed, 38 insertions(+), 29 deletions(-)

diff --git a/hw/msix.c b/hw/msix.c
index 43efbd2..4122395 100644
--- a/hw/msix.c
+++ b/hw/msix.c
@@ -167,35 +167,43 @@ static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,

  {
      int config_offset;
      uint8_t *config;

- uint32_t new_size;

- if (nentries< 1 || nentries> PCI_MSIX_FLAGS_QSIZE + 1)
- return -EINVAL;
- if (bar_size> 0x80000000)
- return -ENOSPC;
-
- /* Add space for MSI-X structures */
- if (!bar_size) {
- new_size = MSIX_PAGE_SIZE;
- } else if (bar_size< MSIX_PAGE_SIZE) {
- bar_size = MSIX_PAGE_SIZE;
- new_size = MSIX_PAGE_SIZE * 2;
- } else {
- new_size = bar_size * 2;
- }
-

 -    pdev->msix_bar_size = new_size;
 -    config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
 -    if (config_offset<  0)
 -        return config_offset;
 -    config = pdev->config + config_offset;
 -
 -    pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
 -    /* Table on top of BAR */
 -    pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
 -    /* Pending bits on top of that */
 -    pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
 -                 bar_nr);
 +    pdev->msix_bar_size = bar_size;
 +
 +    config_offset = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
 +
 +    if (!config_offset) {
 +        uint32_t new_size;
 +
 +        if (nentries<  1 || nentries>  PCI_MSIX_FLAGS_QSIZE + 1)
 +            return -EINVAL;
 +        if (bar_size>  0x80000000)
 +            return -ENOSPC;
 +
 +        /* Add space for MSI-X structures */
 +        if (!bar_size) {
 +            new_size = MSIX_PAGE_SIZE;
 +        } else if (bar_size<  MSIX_PAGE_SIZE) {
 +            bar_size = MSIX_PAGE_SIZE;
 +            new_size = MSIX_PAGE_SIZE * 2;
 +        } else {
 +            new_size = bar_size * 2;
 +        }
 +
 +        pdev->msix_bar_size = new_size;
 +        config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
 +                                           MSIX_CAP_LENGTH);
 +        if (config_offset<  0)
 +            return config_offset;
 +        config = pdev->config + config_offset;
 +
 +        pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
 +        /* Table on top of BAR */
 +        pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
 +        /* Pending bits on top of that */
 +        pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
 +                     bar_nr);
 +    }
      pdev->msix_cap = config_offset;
      /* Make flags bit writeable. */
      pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
 @@ -337,7 +345,8 @@ void msix_mmio_map(PCIDevice *d, int region_num,
          return;
      if (size<= offset)
          return;

- cpu_register_physical_memory(addr + offset, size - offset,
+ cpu_register_physical_memory(addr + offset,
+ MIN(size - offset, MSIX_PAGE_SIZE),

Quoted 8 lines
Quoted 2 lines

This is wrong I think, the table might not fit in a single page. You would need to read table size out of from device config.

That's true, but I was hoping to save that for later since we don't seem to be running into that problem yet. Current device assignment code assumes a single page, and I haven't heard of anyone with a vector table that exceeds that yet. Thanks,

Ok; applied. Please add some warning if the condition happens so the breakage is at least not silent.

Signature · 1 line
error compiling committee.c: too many arguments to function
#5 lore raw

Re: [PATCH 1/2] msix: Allow msix_init on a device with existing MSI-X capability

Quoted 102 lines · Avi Kivity
Quoted 101 lines · Alex Williamson
Quoted 99 lines · Michael S. Tsirkin
Quoted 92 lines · Alex Williamson

To enable common msix support to be used with pass through devices, don't attempt to change the BAR if the device already has an MSI-X capability. This also means we want to pay closer attention to the size when we map the msix table page, as it isn't necessarily covering the entire end of the BAR.

---

  hw/msix.c |   67 +++++++++++++++++++++++++++++++++++--------------------------
  1 files changed, 38 insertions(+), 29 deletions(-)

diff --git a/hw/msix.c b/hw/msix.c
index 43efbd2..4122395 100644
--- a/hw/msix.c
+++ b/hw/msix.c
@@ -167,35 +167,43 @@ static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,

  {
      int config_offset;
      uint8_t *config;

- uint32_t new_size;

- if (nentries< 1 || nentries> PCI_MSIX_FLAGS_QSIZE + 1)
- return -EINVAL;
- if (bar_size> 0x80000000)
- return -ENOSPC;
-
- /* Add space for MSI-X structures */
- if (!bar_size) {
- new_size = MSIX_PAGE_SIZE;
- } else if (bar_size< MSIX_PAGE_SIZE) {
- bar_size = MSIX_PAGE_SIZE;
- new_size = MSIX_PAGE_SIZE * 2;
- } else {
- new_size = bar_size * 2;
- }
-

 -    pdev->msix_bar_size = new_size;
 -    config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
 -    if (config_offset<  0)
 -        return config_offset;
 -    config = pdev->config + config_offset;
 -
 -    pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
 -    /* Table on top of BAR */
 -    pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
 -    /* Pending bits on top of that */
 -    pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
 -                 bar_nr);
 +    pdev->msix_bar_size = bar_size;
 +
 +    config_offset = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
 +
 +    if (!config_offset) {
 +        uint32_t new_size;
 +
 +        if (nentries<  1 || nentries>  PCI_MSIX_FLAGS_QSIZE + 1)
 +            return -EINVAL;
 +        if (bar_size>  0x80000000)
 +            return -ENOSPC;
 +
 +        /* Add space for MSI-X structures */
 +        if (!bar_size) {
 +            new_size = MSIX_PAGE_SIZE;
 +        } else if (bar_size<  MSIX_PAGE_SIZE) {
 +            bar_size = MSIX_PAGE_SIZE;
 +            new_size = MSIX_PAGE_SIZE * 2;
 +        } else {
 +            new_size = bar_size * 2;
 +        }
 +
 +        pdev->msix_bar_size = new_size;
 +        config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
 +                                           MSIX_CAP_LENGTH);
 +        if (config_offset<  0)
 +            return config_offset;
 +        config = pdev->config + config_offset;
 +
 +        pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
 +        /* Table on top of BAR */
 +        pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
 +        /* Pending bits on top of that */
 +        pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
 +                     bar_nr);
 +    }
      pdev->msix_cap = config_offset;
      /* Make flags bit writeable. */
      pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
 @@ -337,7 +345,8 @@ void msix_mmio_map(PCIDevice *d, int region_num,
          return;
      if (size<= offset)
          return;
Quoted 3 lines

- cpu_register_physical_memory(addr + offset, size - offset,
+ cpu_register_physical_memory(addr + offset,
+ MIN(size - offset, MSIX_PAGE_SIZE),

This is wrong I think, the table might not fit in a single page. You would need to read table size out of from device config.

Quoted 8 lines
Quoted 5 lines

That's true, but I was hoping to save that for later since we don't seem to be running into that problem yet. Current device assignment code assumes a single page, and I haven't heard of anyone with a vector table that exceeds that yet. Thanks,

Ok; applied. Please add some warning if the condition happens so the breakage is at least not silent.

Looks like this is already handled before we even get here. msix_init() checks nentries against the max supported in a single page and returns -EINVAL if over. So I think we're safely covered until we add support for more pages. Thanks,

Alex

#6 lore raw

PATCH 2/2 msix: Pull in config.h for CONFIG_KVM

In reply to Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

---

 Makefile.objs   |    2 --
 Makefile.target |    1 +
 hw/msix.c       |    1 +
 3 files changed, 2 insertions(+), 2 deletions(-)
Makefile.objs +0 -2
index ca2d2d0..c097d25 100644
@@ -201,8 +201,6 @@ hw-obj-$(CONFIG_PIIX4) += piix4.o
 # PCI watchdog devices
 hw-obj-y += wdt_i6300esb.o
 
-hw-obj-y += msix.o
-
 # PCI network cards
 hw-obj-y += ne2000.o
 hw-obj-y += eepro100.o
Makefile.target +1 -0
index 347ad6b..63da13b 100644
@@ -185,6 +185,7 @@ obj-y += rwhandler.o
 obj-$(CONFIG_KVM) += kvm.o kvm-all.o
 obj-$(CONFIG_NO_KVM) += kvm-stub.o
 obj-y += memory.o
+obj-y += msix.o
 
 LIBS+=-lz
 
hw/msix.c +1 -0
index 4122395..23256c9 100644
@@ -11,6 +11,7 @@
  * the COPYING file in the top-level directory.
  */
 
+#include "config.h"
 #include "hw.h"
 #include "msix.h"
 #include "pci.h"
#7 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 31 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

---

 Makefile.objs   |    2 --
 Makefile.target |    1 +
 hw/msix.c       |    1 +
 3 files changed, 2 insertions(+), 2 deletions(-)
Makefile.objs +0 -2
index ca2d2d0..c097d25 100644
@@ -201,8 +201,6 @@ hw-obj-$(CONFIG_PIIX4) += piix4.o
 # PCI watchdog devices
 hw-obj-y += wdt_i6300esb.o
 
-hw-obj-y += msix.o
-
 # PCI network cards
 hw-obj-y += ne2000.o
 hw-obj-y += eepro100.o
Makefile.target +0 -0
index 347ad6b..63da13b 100644

Makefile.target obj-y += rwhandler.o · line 188

Quoted 4 lines
Makefile.target +1 -0
@@ -185,6 +185,7 @@ obj-y += rwhandler.o
 obj-$(CONFIG_KVM) += kvm.o kvm-all.o
 obj-$(CONFIG_NO_KVM) += kvm-stub.o
 obj-y += memory.o
+obj-y += msix.o

Oops, memory.c isn't upstream, I'll push it down in my patch queue and send a new one.

Alex

Quoted 15 lines
Makefile.target +0 -0
 LIBS+=-lz
 
hw/msix.c +1 -0
index 4122395..23256c9 100644
@@ -11,6 +11,7 @@
  * the COPYING file in the top-level directory.
  */
 
+#include "config.h"
 #include "hw.h"
 #include "msix.h"
 #include "pci.h"
maintainer
#8 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 2 lines · Alex Williamson

Oops, memory.c isn't upstream, I'll push it down in my patch queue and send a new one.

Neither is kvm_set_irq actually. :) This patch is only needed for qemu-kvm.

BTW, maybe the better solution would be to move the kvm_*_irq* functions from qemu-kvm.c to kvm-all.c, add stubs to kvm-stub.c, and get rid of the #ifdef completely in msix.c

Paolo

#9 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

In reply to Alex Williamson
Quoted 6 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

Why? We just moved it from .target to .objs, see
889e30cc18e21f2091b77267dca8096d7dd34f8b.

Quoted 44 lines

---

 Makefile.objs   |    2 --
 Makefile.target |    1 +
 hw/msix.c       |    1 +
 3 files changed, 2 insertions(+), 2 deletions(-)
Makefile.objs +0 -2
index ca2d2d0..c097d25 100644
@@ -201,8 +201,6 @@ hw-obj-$(CONFIG_PIIX4) += piix4.o
 # PCI watchdog devices
 hw-obj-y += wdt_i6300esb.o
 
-hw-obj-y += msix.o
-
 # PCI network cards
 hw-obj-y += ne2000.o
 hw-obj-y += eepro100.o
Makefile.target +1 -0
index 347ad6b..63da13b 100644
@@ -185,6 +185,7 @@ obj-y += rwhandler.o
 obj-$(CONFIG_KVM) += kvm.o kvm-all.o
 obj-$(CONFIG_NO_KVM) += kvm-stub.o
 obj-y += memory.o
+obj-y += msix.o
 
 LIBS+=-lz
 
hw/msix.c +1 -0
index 4122395..23256c9 100644
@@ -11,6 +11,7 @@
  * the COPYING file in the top-level directory.
  */
 
+#include "config.h"
 #include "hw.h"
 #include "msix.h"
 #include "pci.h"
#10 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 10 lines · Michael S. Tsirkin
Quoted 6 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

Why? We just moved it from .target to .objs, see
889e30cc18e21f2091b77267dca8096d7dd34f8b.

Maybe that's why it used to work. When building in the qemu-kvm.git tree, I'm not getting CONFIG_KVM defined, which means I'm not getting KVM_CAP_IRQCHIP defined, which results in msix_irq_entries not being allocated. Then when I call msix_vector_use, I get a seg fault. Something is broken there. Thanks,

Alex

Quoted 44 lines
Quoted 44 lines

---

 Makefile.objs   |    2 --
 Makefile.target |    1 +
 hw/msix.c       |    1 +
 3 files changed, 2 insertions(+), 2 deletions(-)
Makefile.objs +0 -2
index ca2d2d0..c097d25 100644
@@ -201,8 +201,6 @@ hw-obj-$(CONFIG_PIIX4) += piix4.o
 # PCI watchdog devices
 hw-obj-y += wdt_i6300esb.o
 
-hw-obj-y += msix.o
-
 # PCI network cards
 hw-obj-y += ne2000.o
 hw-obj-y += eepro100.o
Makefile.target +1 -0
index 347ad6b..63da13b 100644
@@ -185,6 +185,7 @@ obj-y += rwhandler.o
 obj-$(CONFIG_KVM) += kvm.o kvm-all.o
 obj-$(CONFIG_NO_KVM) += kvm-stub.o
 obj-y += memory.o
+obj-y += msix.o
 
 LIBS+=-lz
 
hw/msix.c +1 -0
index 4122395..23256c9 100644
@@ -11,6 +11,7 @@
  * the COPYING file in the top-level directory.
  */
 
+#include "config.h"
 #include "hw.h"
 #include "msix.h"
 #include "pci.h"
#11 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 12 lines · Alex Williamson
Quoted 10 lines · Michael S. Tsirkin
Quoted 6 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

Why? We just moved it from .target to .objs, see
889e30cc18e21f2091b77267dca8096d7dd34f8b.

Quoted 7 lines

Maybe that's why it used to work. When building in the qemu-kvm.git tree, I'm not getting CONFIG_KVM defined, which means I'm not getting KVM_CAP_IRQCHIP defined, which results in msix_irq_entries not being allocated. Then when I call msix_vector_use, I get a seg fault. Something is broken there. Thanks,

Alex

This is hopefully fixed in the latest bits.
bd8b215bce453706c3951460cc7e6627ccb90314

Signature · 1 line
MST
#12 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 19 lines · Michael S. Tsirkin
Quoted 18 lines · Alex Williamson
Quoted 10 lines · Michael S. Tsirkin
Quoted 6 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

Why? We just moved it from .target to .objs, see
889e30cc18e21f2091b77267dca8096d7dd34f8b.

Maybe that's why it used to work. When building in the qemu-kvm.git tree, I'm not getting CONFIG_KVM defined, which means I'm not getting KVM_CAP_IRQCHIP defined, which results in msix_irq_entries not being allocated. Then when I call msix_vector_use, I get a seg fault. Something is broken there. Thanks,

Quoted 4 lines
Quoted 1 line

Alex

This is hopefully fixed in the latest bits.
bd8b215bce453706c3951460cc7e6627ccb90314

Nope, my tree includes that. It's not kvm_set_irq, it's kvm_msix_add, which dereferences msix_irq_entries, which is only allocated in msix_init if KVM_CAP_IRQCHIP is defined, which it's not. Maybe you also meant to remove the ifdef from msix_init? I also note there's another in msix_notify. Thanks,

Alex

#13 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 25 lines · Alex Williamson
Quoted 19 lines · Michael S. Tsirkin
Quoted 18 lines · Alex Williamson
Quoted 10 lines · Michael S. Tsirkin
Quoted 6 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined. This requires moving the object over to Makefile.target or else we can't find config-target.h

Why? We just moved it from .target to .objs, see
889e30cc18e21f2091b77267dca8096d7dd34f8b.

Maybe that's why it used to work. When building in the qemu-kvm.git tree, I'm not getting CONFIG_KVM defined, which means I'm not getting KVM_CAP_IRQCHIP defined, which results in msix_irq_entries not being allocated. Then when I call msix_vector_use, I get a seg fault. Something is broken there. Thanks,

Quoted 4 lines
Quoted 1 line

Alex

This is hopefully fixed in the latest bits.
bd8b215bce453706c3951460cc7e6627ccb90314

Quoted 7 lines

Nope, my tree includes that. It's not kvm_set_irq, it's kvm_msix_add, which dereferences msix_irq_entries, which is only allocated in msix_init if KVM_CAP_IRQCHIP is defined, which it's not. Maybe you also meant to remove the ifdef from msix_init? I also note there's another in msix_notify. Thanks,

Alex

Not sure what's wrong:
$git grep KVM_CAP_IRQCHIP origin/master -- hw/msix.c $

http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=blob;f=hw/msix.c;hb=HEAD

also does not show any ifdefs.

Signature · 1 line
MST
#14 lore raw

Re: [PATCH 2/2] msix: Pull in config.h for CONFIG_KVM

Quoted 34 lines · Michael S. Tsirkin
Quoted 25 lines · Alex Williamson
Quoted 19 lines · Michael S. Tsirkin
Quoted 18 lines · Alex Williamson
Quoted 10 lines · Michael S. Tsirkin
Quoted 6 lines · Alex Williamson

We need to pull in config.h or else kvm.h doesn't pull in linux/config.h, which we need if we ever want KVM_CAP_IRQCHIP defined.  This requires moving the object over to Makefile.target or else we can't find config-target.h

Why? We just moved it from .target to .objs, see
889e30cc18e21f2091b77267dca8096d7dd34f8b.

Maybe that's why it used to work.  When building in the qemu-kvm.git tree, I'm not getting CONFIG_KVM defined, which means I'm not getting KVM_CAP_IRQCHIP defined, which results in msix_irq_entries not being allocated.  Then when I call msix_vector_use, I get a seg fault. Something is broken there.  Thanks,

Quoted 4 lines
Quoted 1 line

Alex

This is hopefully fixed in the latest bits.
bd8b215bce453706c3951460cc7e6627ccb90314

Quoted 7 lines

Nope, my tree includes that.  It's not kvm_set_irq, it's kvm_msix_add, which dereferences msix_irq_entries, which is only allocated in msix_init if KVM_CAP_IRQCHIP is defined, which it's not.  Maybe you also meant to remove the ifdef from msix_init?  I also note there's another in msix_notify.  Thanks,

Alex

Quoted 7 lines

Not sure what's wrong:
$git grep KVM_CAP_IRQCHIP origin/master -- hw/msix.c $

http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=blob;f=hw/msix.c;hb=HEAD

also does not show any ifdefs.

Hmm, somehow my tree missed 763a04a920f1098e57ad6b46c91c3e531adc961d Ignore this patch and I'll refresh again. Sorry for the noise.

Alex