History

PCI/MSI: Support allocating virtual MSI interrupts

For NTB devices, we want to be able to trigger MSI interrupts through a memory window. In these cases we may want to use more interrupts than the NTB PCI device has available in its MSI-X table. We allow for this by creating a new 'virtual' interrupt. These interrupts are allocated as usual but are not programmed into the MSI-X table (as there may not be space for them). The MSI address and data will then handled through an NTB MSI library introduced later in this series.

Logan Gunthorpe committed 7 years ago

Commit d7cc6091 parent a944ccc3 files · +62 −9Browse files

Unified · Split

3 files
  1. drivers/pci/msi.c +45 −9
  2. include/linux/msi.h +8 −0
  3. include/linux/pci.h +9 −0
drivers/pci/msi.c+45−9
@@ -192,6 +192,9 @@ static void msi_mask_irq(struct msi_desc *desc, u32 mask, u32 flag) · 191 unmodified lines
192192
193193 static void __iomem *pci_msix_desc_addr(struct msi_desc *desc)
194194 {
195+ if (desc->msi_attrib.is_virtual)
196+ return NULL;
197+
195198 return desc->mask_base +
196199 desc->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE;
197200 }
@@ -206,14 +209,19 @@ static void __iomem *pci_msix_desc_addr(struct msi_desc *desc) · 8 unmodified lines
206209 u32 __pci_msix_desc_mask_irq(struct msi_desc *desc, u32 flag)
207210 {
208211 u32 mask_bits = desc->masked;
212+ void __iomem *desc_addr;
209213
210214 if (pci_msi_ignore_mask)
211215 return 0;
216+ desc_addr = pci_msix_desc_addr(desc);
217+ if (!desc_addr)
218+ return 0;
212219
213220 mask_bits &= ~PCI_MSIX_ENTRY_CTRL_MASKBIT;
214221 if (flag)
215222 mask_bits |= PCI_MSIX_ENTRY_CTRL_MASKBIT;
216− writel(mask_bits, pci_msix_desc_addr(desc) + PCI_MSIX_ENTRY_VECTOR_CTRL);
223+
224+ writel(mask_bits, desc_addr + PCI_MSIX_ENTRY_VECTOR_CTRL);
217225
218226 return mask_bits;
219227 }
@@ -273,6 +281,11 @@ void __pci_read_msi_msg(struct msi_desc *entry, struct msi_msg *msg) · 53 unmodified lines
273281 if (entry->msi_attrib.is_msix) {
274282 void __iomem *base = pci_msix_desc_addr(entry);
275283
284+ if (!base) {
285+ WARN_ON(1);
286+ return;
287+ }
288+
276289 msg->address_lo = readl(base + PCI_MSIX_ENTRY_LOWER_ADDR);
277290 msg->address_hi = readl(base + PCI_MSIX_ENTRY_UPPER_ADDR);
278291 msg->data = readl(base + PCI_MSIX_ENTRY_DATA);
@@ -303,6 +316,9 @@ void __pci_write_msi_msg(struct msi_desc *entry, struct msi_msg *msg) · 24 unmodified lines
303316 } else if (entry->msi_attrib.is_msix) {
304317 void __iomem *base = pci_msix_desc_addr(entry);
305318
319+ if (!base)
320+ goto skip;
321+
306322 writel(msg->address_lo, base + PCI_MSIX_ENTRY_LOWER_ADDR);
307323 writel(msg->address_hi, base + PCI_MSIX_ENTRY_UPPER_ADDR);
308324 writel(msg->data, base + PCI_MSIX_ENTRY_DATA);
@@ -327,7 +343,13 @@ void __pci_write_msi_msg(struct msi_desc *entry, struct msi_msg *msg) · 18 unmodified lines
327343 msg->data);
328344 }
329345 }
346+
347+skip:
330348 entry->msg = *msg;
349+
350+ if (entry->write_msi_msg)
351+ entry->write_msi_msg(entry, entry->write_msi_msg_data);
352+
331353 }
332354
333355 void pci_write_msi_msg(unsigned int irq, struct msi_msg *msg)
@@ -550,6 +572,7 @@ msi_setup_entry(struct pci_dev *dev, int nvec, struct irq_affinity *affd) · 216 unmodified lines
550572
551573 entry->msi_attrib.is_msix = 0;
552574 entry->msi_attrib.is_64 = !!(control & PCI_MSI_FLAGS_64BIT);
575+ entry->msi_attrib.is_virtual = 0;
553576 entry->msi_attrib.entry_nr = 0;
554577 entry->msi_attrib.maskbit = !!(control & PCI_MSI_FLAGS_MASKBIT);
555578 entry->msi_attrib.default_irq = dev->irq; /* Save IOAPIC IRQ */
@@ -674,6 +697,7 @@ static int msix_setup_entries(struct pci_dev *dev, void __iomem *base, · 118 unmodified lines
674697 struct irq_affinity_desc *curmsk, *masks = NULL;
675698 struct msi_desc *entry;
676699 int ret, i;
700+ int vec_count = pci_msix_vec_count(dev);
677701
678702 if (affd)
679703 masks = irq_create_affinity_masks(nvec, affd);
@@ -696,6 +720,10 @@ static int msix_setup_entries(struct pci_dev *dev, void __iomem *base, · 16 unmodified lines
696720 entry->msi_attrib.entry_nr = entries[i].entry;
697721 else
698722 entry->msi_attrib.entry_nr = i;
723+
724+ entry->msi_attrib.is_virtual =
725+ entry->msi_attrib.entry_nr >= vec_count;
726+
699727 entry->msi_attrib.default_irq = dev->irq;
700728 entry->mask_base = base;
701729
@@ -714,12 +742,19 @@ static void msix_program_entries(struct pci_dev *dev, · 12 unmodified lines
714742 {
715743 struct msi_desc *entry;
716744 int i = 0;
745+ void __iomem *desc_addr;
717746
718747 for_each_pci_msi_entry(entry, dev) {
719748 if (entries)
720749 entries[i++].vector = entry->irq;
721− entry->masked = readl(pci_msix_desc_addr(entry) +
722− PCI_MSIX_ENTRY_VECTOR_CTRL);
750+
751+ desc_addr = pci_msix_desc_addr(entry);
752+ if (desc_addr)
753+ entry->masked = readl(desc_addr +
754+ PCI_MSIX_ENTRY_VECTOR_CTRL);
755+ else
756+ entry->masked = 0;
757+
723758 msix_mask_irq(entry, 1);
724759 }
725760 }
@@ -932,7 +967,7 @@ int pci_msix_vec_count(struct pci_dev *dev) · 206 unmodified lines
932967 EXPORT_SYMBOL(pci_msix_vec_count);
933968
934969 static int __pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries,
935− int nvec, struct irq_affinity *affd)
970+ int nvec, struct irq_affinity *affd, int flags)
936971 {
937972 int nr_entries;
938973 int i, j;
@@ -943,7 +978,7 @@ static int __pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries, · 4 unmodified lines
943978 nr_entries = pci_msix_vec_count(dev);
944979 if (nr_entries < 0)
945980 return nr_entries;
946− if (nvec > nr_entries)
981+ if (nvec > nr_entries && !(flags & PCI_IRQ_VIRTUAL))
947982 return nr_entries;
948983
949984 if (entries) {
@@ -1079,7 +1114,8 @@ EXPORT_SYMBOL(pci_enable_msi); · 129 unmodified lines
10791114
10801115 static int __pci_enable_msix_range(struct pci_dev *dev,
10811116 struct msix_entry *entries, int minvec,
1082− int maxvec, struct irq_affinity *affd)
1117+ int maxvec, struct irq_affinity *affd,
1118+ int flags)
10831119 {
10841120 int rc, nvec = maxvec;
10851121
@@ -1096,7 +1132,7 @@ static int __pci_enable_msix_range(struct pci_dev *dev, · 10 unmodified lines
10961132 return -ENOSPC;
10971133 }
10981134
1099− rc = __pci_enable_msix(dev, entries, nvec, affd);
1135+ rc = __pci_enable_msix(dev, entries, nvec, affd, flags);
11001136 if (rc == 0)
11011137 return nvec;
11021138
@@ -1127,7 +1163,7 @@ static int __pci_enable_msix_range(struct pci_dev *dev, · 24 unmodified lines
11271163 int pci_enable_msix_range(struct pci_dev *dev, struct msix_entry *entries,
11281164 int minvec, int maxvec)
11291165 {
1130− return __pci_enable_msix_range(dev, entries, minvec, maxvec, NULL);
1166+ return __pci_enable_msix_range(dev, entries, minvec, maxvec, NULL, 0);
11311167 }
11321168 EXPORT_SYMBOL(pci_enable_msix_range);
11331169
@@ -1167,7 +1203,7 @@ int pci_alloc_irq_vectors_affinity(struct pci_dev *dev, unsigned int min_vecs, · 33 unmodified lines
11671203
11681204 if (flags & PCI_IRQ_MSIX) {
11691205 msix_vecs = __pci_enable_msix_range(dev, NULL, min_vecs,
1170− max_vecs, affd);
1206+ max_vecs, affd, flags);
11711207 if (msix_vecs > 0)
11721208 return msix_vecs;
11731209 }
include/linux/msi.h+8−0
@@ -64,6 +64,10 @@ struct ti_sci_inta_msi_desc { · 63 unmodified lines
6464 * @msg: The last set MSI message cached for reuse
6565 * @affinity: Optional pointer to a cpu affinity mask for this descriptor
6666 *
67+ * @write_msi_msg: Callback that may be called when the MSI message
68+ * address or data changes
69+ * @write_msi_msg_data: Data parameter for the callback.
70+ *
6771 * @masked: [PCI MSI/X] Mask bits
6872 * @is_msix: [PCI MSI/X] True if MSI-X
6973 * @multiple: [PCI MSI/X] log2 num of messages allocated
@@ -90,6 +94,9 @@ struct msi_desc { · 20 unmodified lines
9094 const void *iommu_cookie;
9195 #endif
9296
97+ void (*write_msi_msg)(struct msi_desc *entry, void *data);
98+ void *write_msi_msg_data;
99+
93100 union {
94101 /* PCI MSI/X specific data */
95102 struct {
@@ -100,6 +107,7 @@ struct msi_desc { · 4 unmodified lines
100107 u8 multi_cap : 3;
101108 u8 maskbit : 1;
102109 u8 is_64 : 1;
110+ u8 is_virtual : 1;
103111 u16 entry_nr;
104112 unsigned default_irq;
105113 } msi_attrib;
include/linux/pci.h+9−0
@@ -1362,6 +1362,15 @@ int pci_set_vga_state(struct pci_dev *pdev, bool decode, · 1361 unmodified lines
13621362 #define PCI_IRQ_MSI (1 << 1) /* Allow MSI interrupts */
13631363 #define PCI_IRQ_MSIX (1 << 2) /* Allow MSI-X interrupts */
13641364 #define PCI_IRQ_AFFINITY (1 << 3) /* Auto-assign affinity */
1365+
1366+/*
1367+ * Virtual interrupts allow for more interrupts to be allocated
1368+ * than the device has interrupts for. These are not programmed
1369+ * into the device's MSI-X table and must be handled by some
1370+ * other driver means.
1371+ */
1372+#define PCI_IRQ_VIRTUAL (1 << 4)
1373+
13651374 #define PCI_IRQ_ALL_TYPES \
13661375 (PCI_IRQ_LEGACY | PCI_IRQ_MSI | PCI_IRQ_MSIX)
13671376

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t