From: Don Dugger <ddugger@redhat.com>
Date: Fri, 18 Dec 2009 20:46:38 -0500
Subject: Revert: [xen] fix msi-x table fixmap allocation
Message-id: <200912182046.nBIKkcXK018219@sobek.n0ano.com>
Patchwork-id: 22183
O-Subject: [RHEL5.5 PATCH 3/3 V2] BZ 537734: MSI-X mask bit acceleration
Bugzilla: 537734

Reverting due to xen hv abi breakage:

    Message-id: <200911302157.nAULvF8p008702@sobek.n0ano.com>
    Patchwork-id: 21558
    O-Subject: [RHEL5.5 PATCH 1/3] BZ 537734: xen: fix msix table fixmap allocation
    Bugzilla: 537734
    RH-Acked-by: Christopher Lalancette <clalance@redhat.com>
    RH-Acked-by: Prarit Bhargava <prarit@redhat.com>
    RH-Acked-by: Miroslav Rezanina <mrezanin@redhat.com>

    backporting xen-unstable changeset 19216

    changeset:   19216:2e6de0f50f3f
        passthrough: fix MSI-X table fixmap allocation

        Currently, msix table pages are allocated a fixmap page per vector,
        the available fixmap pages will be depleted when assigning devices
        with large number of vectors.  This patch fixes it, and a bug that
        prevents cross-page MSI-X table from working properly

        It now allocates msix table fixmap pages per device, if the table
        entries of two msix vectors share the same page, it will only be
        mapped to fixmap once. A ref count is maintained so that it can
        be unmapped when all the vectors are freed.

        Also changes the meaning of msi_desc->mask_base from the va of msix
        table start to the va of the target entry. The former one is currently
        buggy (it always maps the first page but msix can support up to 2048
        entries) and can't handle separately allocated pages.

        Signed-off-by: Qing He <qing.he@intel.com>

    Signed-off-by: Qing He <qing.he@intel.com>
    Signed-off-by: Don Dugger <donald.d.dugger.com>

Signed-off-by: Don Dugger <donald.d.dugger@intel.com>
---
 arch/x86/msi.c            |  109 ++++++++++++--------------------------------
 drivers/passthrough/pci.c |    1 -
 include/asm-x86/fixmap.h  |    2 +-
 include/asm-x86/msi.h     |    6 +-
 include/xen/pci.h         |    6 ---
 5 files changed, 34 insertions(+), 90 deletions(-)

Signed-off-by: Jarod Wilson <jarod@redhat.com>

diff --git a/arch/x86/msi.c b/arch/x86/msi.c
index 4c13ad3..fdca245 100644
--- a/arch/x86/msi.c
+++ b/arch/x86/msi.c
@@ -29,17 +29,17 @@
 
 /* bitmap indicate which fixed map is free */
 DEFINE_SPINLOCK(msix_fixmap_lock);
-DECLARE_BITMAP(msix_fixmap_pages, FIX_MSIX_MAX_PAGES);
+DECLARE_BITMAP(msix_fixmap_pages, MAX_MSIX_PAGES);
 
 static int msix_fixmap_alloc(void)
 {
-    int i, rc = -ENOMEM;
+    int i, rc = -1;
 
     spin_lock(&msix_fixmap_lock);
-    for ( i = 0; i < FIX_MSIX_MAX_PAGES; i++ )
+    for ( i = 0; i < MAX_MSIX_PAGES; i++ )
         if ( !test_bit(i, &msix_fixmap_pages) )
             break;
-    if ( i == FIX_MSIX_MAX_PAGES )
+    if ( i == MAX_MSIX_PAGES )
         goto out;
     rc = FIX_MSIX_IO_RESERV_BASE + i;
     set_bit(i, &msix_fixmap_pages);
@@ -51,66 +51,8 @@ static int msix_fixmap_alloc(void)
 
 static void msix_fixmap_free(int idx)
 {
-    spin_lock(&msix_fixmap_lock);
     if ( idx >= FIX_MSIX_IO_RESERV_BASE )
         clear_bit(idx - FIX_MSIX_IO_RESERV_BASE, &msix_fixmap_pages);
-    spin_unlock(&msix_fixmap_lock);
-}
-
-static int msix_get_fixmap(struct pci_dev *dev, unsigned long table_paddr,
-                           unsigned long entry_paddr)
-{
-    int nr_page, idx;
-
-    nr_page = (entry_paddr >> PAGE_SHIFT) - (table_paddr >> PAGE_SHIFT);
-
-    if ( nr_page < 0 || nr_page >= MAX_MSIX_TABLE_PAGES )
-        return -EINVAL;
-
-    spin_lock(&dev->msix_table_lock);
-    if ( dev->msix_table_refcnt[nr_page]++ == 0 )
-    {
-        idx = msix_fixmap_alloc();
-        if ( idx < 0 )
-        {
-            dev->msix_table_refcnt[nr_page]--;
-            goto out;
-        }
-        set_fixmap_nocache(idx, entry_paddr);
-        dev->msix_table_idx[nr_page] = idx;
-    }
-    else
-        idx = dev->msix_table_idx[nr_page];
-
- out:
-    spin_unlock(&dev->msix_table_lock);
-    return idx;
-}
-
-static void msix_put_fixmap(struct pci_dev *dev, int idx)
-{
-    int i;
-    unsigned long start;
-
-    spin_lock(&dev->msix_table_lock);
-    for ( i = 0; i < MAX_MSIX_TABLE_PAGES; i++ )
-    {
-        if ( dev->msix_table_idx[i] == idx )
-            break;
-    }
-    if ( i == MAX_MSIX_TABLE_PAGES )
-        goto out;
-
-    if ( --dev->msix_table_refcnt[i] == 0 )
-    {
-        start = fix_to_virt(idx);
-        destroy_xen_mappings(start, start + PAGE_SIZE);
-        msix_fixmap_free(idx);
-        dev->msix_table_idx[i] = 0;
-    }
-
- out:
-    spin_unlock(&dev->msix_table_lock);
 }
 
 /*
@@ -180,7 +122,8 @@ static void read_msi_msg(struct msi_desc *entry, struct msi_msg *msg)
     case PCI_CAP_ID_MSIX:
     {
         void __iomem *base;
-        base = entry->mask_base;
+        base = entry->mask_base +
+            entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE;
 
         msg->address_lo = readl(base + PCI_MSIX_ENTRY_LOWER_ADDR_OFFSET);
         msg->address_hi = readl(base + PCI_MSIX_ENTRY_UPPER_ADDR_OFFSET);
@@ -256,7 +199,8 @@ static void write_msi_msg(struct msi_desc *entry, struct msi_msg *msg)
     case PCI_CAP_ID_MSIX:
     {
         void __iomem *base;
-        base = entry->mask_base;
+        base = entry->mask_base +
+            entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE;
 
         writel(msg->address_lo,
                base + PCI_MSIX_ENTRY_LOWER_ADDR_OFFSET);
@@ -344,7 +288,8 @@ static void msix_flush_writes(unsigned int vector)
         break;
     case PCI_CAP_ID_MSIX:
     {
-        int offset = PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET;
+        int offset = entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE +
+            PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET;
         readl(entry->mask_base + offset);
         break;
     }
@@ -385,7 +330,8 @@ static void msi_set_mask_bit(unsigned int vector, int flag)
         break;
     case PCI_CAP_ID_MSIX:
     {
-        int offset = PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET;
+        int offset = entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE +
+            PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET;
         writel(flag, entry->mask_base + offset);
         readl(entry->mask_base + offset);
         break;
@@ -446,10 +392,13 @@ int msi_free_vector(struct msi_desc *entry)
     {
         unsigned long start;
 
-        writel(1, entry->mask_base + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
+        writel(1, entry->mask_base + entry->msi_attrib.entry_nr
+               * PCI_MSIX_ENTRY_SIZE
+               + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
 
         start = (unsigned long)entry->mask_base & ~(PAGE_SIZE - 1);
-        msix_put_fixmap(entry->dev, virt_to_fix(start));
+        msix_fixmap_free(virt_to_fix(start));
+        destroy_xen_mappings(start, start + PAGE_SIZE);
     }
     list_del(&entry->list);
     xfree(entry);
@@ -551,8 +500,8 @@ static int msix_capability_init(struct pci_dev *dev,
     struct msi_desc *entry;
     int pos;
     u16 control;
-    unsigned long table_paddr, entry_paddr;
-    u32 table_offset, entry_offset;
+    unsigned long phys_addr;
+    u32 table_offset;
     u8 bir;
     void __iomem *base;
     int idx;
@@ -576,17 +525,15 @@ static int msix_capability_init(struct pci_dev *dev,
     table_offset = pci_conf_read32(bus, slot, func, msix_table_offset_reg(pos));
     bir = (u8)(table_offset & PCI_MSIX_FLAGS_BIRMASK);
     table_offset &= ~PCI_MSIX_FLAGS_BIRMASK;
-    entry_offset = msi->entry_nr * PCI_MSIX_ENTRY_SIZE;
-
-    table_paddr = msi->table_base + table_offset;
-    entry_paddr = table_paddr + entry_offset;
-    idx = msix_get_fixmap(dev, table_paddr, entry_paddr);
+    phys_addr = msi->table_base + table_offset;
+    idx = msix_fixmap_alloc();
     if ( idx < 0 )
     {
         xfree(entry);
-        return idx;
+        return -ENOMEM;
     }
-    base = (void *)(fix_to_virt(idx) + (entry_paddr & ((1UL << PAGE_SHIFT) - 1)));
+    set_fixmap_nocache(idx, phys_addr);
+    base = (void *)(fix_to_virt(idx) + (phys_addr & ((1UL << PAGE_SHIFT) - 1)));
 
     entry->msi_attrib.type = PCI_CAP_ID_MSIX;
     entry->msi_attrib.is_64 = 1;
@@ -601,7 +548,9 @@ static int msix_capability_init(struct pci_dev *dev,
     list_add_tail(&entry->list, &dev->msi_list);
 
     /* Mask interrupt here */
-    writel(1, entry->mask_base + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
+    writel(1, entry->mask_base + entry->msi_attrib.entry_nr
+                * PCI_MSIX_ENTRY_SIZE
+                + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
 
     *desc = entry;
     /* Restore MSI-X enabled bits */
@@ -726,7 +675,9 @@ static void __pci_disable_msix(struct msi_desc *entry)
 
     BUG_ON(list_empty(&dev->msi_list));
 
-    writel(1, entry->mask_base + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
+    writel(1, entry->mask_base + entry->msi_attrib.entry_nr
+      * PCI_MSIX_ENTRY_SIZE
+      + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
 
     pci_conf_write16(bus, slot, func, msix_control_reg(pos), control);
 }
diff --git a/drivers/passthrough/pci.c b/drivers/passthrough/pci.c
index 98ae208..c52b7e2 100644
--- a/drivers/passthrough/pci.c
+++ b/drivers/passthrough/pci.c
@@ -48,7 +48,6 @@ struct pci_dev *alloc_pdev(u8 bus, u8 devfn)
     pdev->domain = NULL;
     INIT_LIST_HEAD(&pdev->msi_list);
     list_add(&pdev->alldevs_list, &alldevs_list);
-    spin_lock_init(&pdev->msix_table_lock);
 
     return pdev;
 }
diff --git a/include/asm-x86/fixmap.h b/include/asm-x86/fixmap.h
index 910a462..4d1005e 100644
--- a/include/asm-x86/fixmap.h
+++ b/include/asm-x86/fixmap.h
@@ -48,7 +48,7 @@ enum fixed_addresses {
     FIX_IOMMU_MMIO_BASE_0,
     FIX_IOMMU_MMIO_END = FIX_IOMMU_MMIO_BASE_0 + IOMMU_PAGES-1,
     FIX_MSIX_IO_RESERV_BASE,
-    FIX_MSIX_IO_RESERV_END = FIX_MSIX_IO_RESERV_BASE + FIX_MSIX_MAX_PAGES -1,
+    FIX_MSIX_IO_RESERV_END = FIX_MSIX_IO_RESERV_BASE + MAX_MSIX_PAGES -1,
     __end_of_fixed_addresses
 };
 
diff --git a/include/asm-x86/msi.h b/include/asm-x86/msi.h
index a5f67e1..1f18b7c 100644
--- a/include/asm-x86/msi.h
+++ b/include/asm-x86/msi.h
@@ -49,9 +49,9 @@
 
 /* MAX fixed pages reserved for mapping MSIX tables. */
 #if defined(__x86_64__)
-#define FIX_MSIX_MAX_PAGES              512
+#define MAX_MSIX_PAGES              512
 #else
-#define FIX_MSIX_MAX_PAGES              32
+#define MAX_MSIX_PAGES              32
 #endif
 
 struct msi_info {
@@ -93,7 +93,7 @@ struct msi_desc {
 
 	struct list_head list;
 
-	void __iomem *mask_base;        /* va for the entry in mask table */
+	void __iomem *mask_base;
 	struct pci_dev *dev;
 	int vector;
 
diff --git a/include/xen/pci.h b/include/xen/pci.h
index 0fd22f0..e7c98c5 100644
--- a/include/xen/pci.h
+++ b/include/xen/pci.h
@@ -29,16 +29,10 @@
 #define PCI_BDF(b,d,f)  ((((b) & 0xff) << 8) | PCI_DEVFN(d,f))
 #define PCI_BDF2(b,df)  ((((b) & 0xff) << 8) | ((df) & 0xff))
 
-#define MAX_MSIX_TABLE_PAGES    8    /* 2048 entries */
 struct pci_dev {
     struct list_head alldevs_list;
     struct list_head domain_list;
-
     struct list_head msi_list;
-    int msix_table_refcnt[MAX_MSIX_TABLE_PAGES];
-    int msix_table_idx[MAX_MSIX_TABLE_PAGES];
-    spinlock_t msix_table_lock;
-
     struct domain *domain;
     const u8 bus;
     const u8 devfn;