IOMMU, vtd and iotlb flush rework (v2)

IOMMU, vtd and iotlb flush rework (v2)

[Jean Guyader]

In one of my previous email I detailed a bug I was seeing when passing
through a Intel GPU on a guest that has more that 4G or RAM.

Allen suggested that I go for the Plan B but after a discussion with Tim
we agreed that Plan B was way to disruptive in term of code change.

This patch series implements Plan A.

http://xen.1045712.n5.nabble.com/VTD-Intel-iommu-IOTLB-flush-really-slow-td4952866.html

Changes between v1 and v2:
        - Move size in struct xen_add_to_physmap in padding between .domid and .space.
        - Store iommu_dont_flush per cpu
        - Change the code in hvmloader to relocate by batch of 64K, .size is now 16 bits.

Jean Guyader (6):
      vtd: Refactor iotlb flush code
      iommu: Introduce iommu_flush and iommu_flush_all.
      add_to_physmap: Move the code for XENMEM_add_to_physmap.
      mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range
      hvmloader: Change memory relocation loop when overlap with PCI hole.
      Introduce per cpu flag (iommu_dont_flush_iotlb) to avoid unnecessary iotlb flush
 
 tools/firmware/hvmloader/pci.c      |   20 +++-
 xen/arch/x86/mm.c                   |  203 +++++++++++++++++++++--------------
 xen/drivers/passthrough/iommu.c     |   25 +++++
 xen/drivers/passthrough/vtd/iommu.c |  100 ++++++++++--------
 xen/include/public/memory.h         |    5 +-
 xen/include/xen/iommu.h             |    7 ++
 6 files changed, 230 insertions(+), 130 deletions(-)

Jean

[PATCH 1/6] vtd: Refactor iotlb flush code

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 279 bytes --]


Factorize the iotlb flush code from map_page and unmap_page into
it's own function.

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/drivers/passthrough/vtd/iommu.c |   86 +++++++++++++++++-----------------
 1 files changed, 43 insertions(+), 43 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-vtd-Refactor-iotlb-flush-code.patch --]
[-- Type: text/x-patch; name="0001-vtd-Refactor-iotlb-flush-code.patch", Size: 4122 bytes --]

diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 7717ab4..5a5b6be 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -578,16 +578,53 @@ static void iommu_flush_all(void)
     }
 }
 
+static void __intel_iommu_iotlb_flush(struct domain *d, unsigned long gfn,
+        int dma_old_pte_present, unsigned int page_count)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(d);
+    struct acpi_drhd_unit *drhd;
+    struct iommu *iommu;
+    int flush_dev_iotlb;
+    int iommu_domid;
+
+    /*
+     * No need pcideves_lock here because we have flush
+     * when assign/deassign device
+     */
+    for_each_drhd_unit ( drhd )
+    {
+        iommu = drhd->iommu;
+
+        if ( !test_bit(iommu->index, &hd->iommu_bitmap) )
+            continue;
+
+        flush_dev_iotlb = find_ats_dev_drhd(iommu) ? 1 : 0;
+        iommu_domid= domain_iommu_domid(d, iommu);
+        if ( iommu_domid == -1 )
+            continue;
+
+        if ( page_count > 1 || gfn == -1 )
+        {
+            if ( iommu_flush_iotlb_dsi(iommu, iommu_domid,
+                        0, flush_dev_iotlb) )
+                iommu_flush_write_buffer(iommu);
+        }
+        else
+        {
+            if ( iommu_flush_iotlb_psi(iommu, iommu_domid,
+                        (paddr_t)gfn << PAGE_SHIFT_4K, 0,
+                        !dma_old_pte_present, flush_dev_iotlb) )
+                iommu_flush_write_buffer(iommu);
+        }
+    }
+}
+
 /* clear one page's page table */
 static void dma_pte_clear_one(struct domain *domain, u64 addr)
 {
     struct hvm_iommu *hd = domain_hvm_iommu(domain);
-    struct acpi_drhd_unit *drhd;
-    struct iommu *iommu;
     struct dma_pte *page = NULL, *pte = NULL;
     u64 pg_maddr;
-    int flush_dev_iotlb;
-    int iommu_domid;
     struct mapped_rmrr *mrmrr;
 
     spin_lock(&hd->mapping_lock);
@@ -613,21 +650,7 @@ static void dma_pte_clear_one(struct domain *domain, u64 addr)
     spin_unlock(&hd->mapping_lock);
     iommu_flush_cache_entry(pte, sizeof(struct dma_pte));
 
-    /* No need pcidevs_lock here since do that on assign/deassign device*/
-    for_each_drhd_unit ( drhd )
-    {
-        iommu = drhd->iommu;
-        if ( test_bit(iommu->index, &hd->iommu_bitmap) )
-        {
-            flush_dev_iotlb = find_ats_dev_drhd(iommu) ? 1 : 0;
-            iommu_domid= domain_iommu_domid(domain, iommu);
-            if ( iommu_domid == -1 )
-                continue;
-            if ( iommu_flush_iotlb_psi(iommu, iommu_domid, addr,
-                                       0, 0, flush_dev_iotlb) )
-                iommu_flush_write_buffer(iommu);
-        }
-    }
+    __intel_iommu_iotlb_flush(domain, addr >> PAGE_SHIFT_4K , 0, 1);
 
     unmap_vtd_domain_page(page);
 
@@ -1677,12 +1700,8 @@ static int intel_iommu_map_page(
     unsigned int flags)
 {
     struct hvm_iommu *hd = domain_hvm_iommu(d);
-    struct acpi_drhd_unit *drhd;
-    struct iommu *iommu;
     struct dma_pte *page = NULL, *pte = NULL, old, new = { 0 };
     u64 pg_maddr;
-    int flush_dev_iotlb;
-    int iommu_domid;
 
     /* Do nothing if VT-d shares EPT page table */
     if ( iommu_use_hap_pt(d) )
@@ -1724,26 +1743,7 @@ static int intel_iommu_map_page(
     spin_unlock(&hd->mapping_lock);
     unmap_vtd_domain_page(page);
 
-    /*
-     * No need pcideves_lock here because we have flush
-     * when assign/deassign device
-     */
-    for_each_drhd_unit ( drhd )
-    {
-        iommu = drhd->iommu;
-
-        if ( !test_bit(iommu->index, &hd->iommu_bitmap) )
-            continue;
-
-        flush_dev_iotlb = find_ats_dev_drhd(iommu) ? 1 : 0;
-        iommu_domid= domain_iommu_domid(d, iommu);
-        if ( iommu_domid == -1 )
-            continue;
-        if ( iommu_flush_iotlb_psi(iommu, iommu_domid,
-                                   (paddr_t)gfn << PAGE_SHIFT_4K, 0,
-                                   !dma_pte_present(old), flush_dev_iotlb) )
-            iommu_flush_write_buffer(iommu);
-    }
+    __intel_iommu_iotlb_flush(d, gfn, dma_pte_present(old), 1);
 
     return 0;
 }

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 286 bytes --]


Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
 xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
 xen/include/xen/iommu.h             |    5 +++++
 3 files changed, 37 insertions(+), 0 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0002-iommu-Introduce-iommu_flush-and-iommu_flush_all.patch --]
[-- Type: text/x-patch; name="0002-iommu-Introduce-iommu_flush-and-iommu_flush_all.patch", Size: 2742 bytes --]

diff --git a/xen/drivers/passthrough/iommu.c b/xen/drivers/passthrough/iommu.c
index cd6174d..9c62861 100644
--- a/xen/drivers/passthrough/iommu.c
+++ b/xen/drivers/passthrough/iommu.c
@@ -301,6 +301,26 @@ int iommu_unmap_page(struct domain *d, unsigned long gfn)
     return hd->platform_ops->unmap_page(d, gfn);
 }
 
+void iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(d);
+
+    if ( !iommu_enabled || !hd->platform_ops || !hd->platform_ops->iotlb_flush )
+        return;
+
+    hd->platform_ops->iotlb_flush(d, gfn, page_count);
+}
+
+void iommu_iotlb_flush_all(struct domain *d)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(d);
+
+    if ( !iommu_enabled || !hd->platform_ops )
+        return;
+
+    hd->platform_ops->iotlb_flush_all(d);
+}
+
 /* caller should hold the pcidevs_lock */
 int deassign_device(struct domain *d, u16 seg, u8 bus, u8 devfn)
 {
diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 5a5b6be..7ec9541 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -619,6 +619,16 @@ static void __intel_iommu_iotlb_flush(struct domain *d, unsigned long gfn,
     }
 }
 
+static void intel_iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count)
+{
+    __intel_iommu_iotlb_flush(d, gfn, 1, page_count);
+}
+
+static void intel_iommu_iotlb_flush_all(struct domain *d)
+{
+    __intel_iommu_iotlb_flush(d, 0, 0, 0);
+}
+
 /* clear one page's page table */
 static void dma_pte_clear_one(struct domain *domain, u64 addr)
 {
@@ -2329,6 +2339,8 @@ const struct iommu_ops intel_iommu_ops = {
     .resume = vtd_resume,
     .share_p2m = iommu_set_pgd,
     .crash_shutdown = vtd_crash_shutdown,
+    .iotlb_flush = intel_iommu_iotlb_flush,
+    .iotlb_flush_all = intel_iommu_iotlb_flush_all,
 };
 
 /*
diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
index 837e60d..a1034df 100644
--- a/xen/include/xen/iommu.h
+++ b/xen/include/xen/iommu.h
@@ -139,6 +139,8 @@ struct iommu_ops {
     void (*resume)(void);
     void (*share_p2m)(struct domain *d);
     void (*crash_shutdown)(void);
+    void (*iotlb_flush)(struct domain *d, unsigned long gfn, unsigned int page_count);
+    void (*iotlb_flush_all)(struct domain *d);
 };
 
 void iommu_update_ire_from_apic(unsigned int apic, unsigned int reg, unsigned int value);
@@ -155,4 +157,7 @@ void iommu_share_p2m_table(struct domain *d);
 
 int iommu_do_domctl(struct xen_domctl *, XEN_GUEST_HANDLE(xen_domctl_t));
 
+void iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count);
+void iommu_iotlb_flush_all(struct domain *d);
+
 #endif /* _IOMMU_H_ */

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 3/6] add_to_physmap: Move the code for XENMEM_add_to_physmap.

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 293 bytes --]


Move the code for the XENMEM_add_to_physmap case into it's own
function (xenmem_add_to_physmap).

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/arch/x86/mm.c |  188 ++++++++++++++++++++++++++++-------------------------
 1 files changed, 99 insertions(+), 89 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0003-add_to_physmap-Move-the-code-for-XENMEM_add_to_physm.patch --]
[-- Type: text/x-patch; name="0003-add_to_physmap-Move-the-code-for-XENMEM_add_to_physm.patch", Size: 7065 bytes --]

diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index acc1f34..f75011e 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -4592,119 +4592,129 @@ static int handle_iomem_range(unsigned long s, unsigned long e, void *p)
     return 0;
 }
 
-long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
+static int xenmem_add_to_physmap(struct domain *d, struct xen_add_to_physmap xatp)
 {
     struct page_info *page = NULL;
+    unsigned long prev_mfn, mfn = 0, gpfn;
     int rc;
 
-    switch ( op )
-    {
-    case XENMEM_add_to_physmap:
+    switch ( xatp.space )
     {
-        struct xen_add_to_physmap xatp;
-        unsigned long prev_mfn, mfn = 0, gpfn;
-        struct domain *d;
-
-        if ( copy_from_guest(&xatp, arg, 1) )
-            return -EFAULT;
+    case XENMAPSPACE_shared_info:
+        if ( xatp.idx == 0 )
+            mfn = virt_to_mfn(d->shared_info);
+        break;
+    case XENMAPSPACE_grant_table:
+        spin_lock(&d->grant_table->lock);
 
-        rc = rcu_lock_target_domain_by_id(xatp.domid, &d);
-        if ( rc != 0 )
-            return rc;
+        if ( d->grant_table->gt_version == 0 )
+            d->grant_table->gt_version = 1;
 
-        if ( xsm_add_to_physmap(current->domain, d) )
+        if ( d->grant_table->gt_version == 2 &&
+             (xatp.idx & XENMAPIDX_grant_table_status) )
         {
-            rcu_unlock_domain(d);
-            return -EPERM;
+            xatp.idx &= ~XENMAPIDX_grant_table_status;
+            if ( xatp.idx < nr_status_frames(d->grant_table) )
+                mfn = virt_to_mfn(d->grant_table->status[xatp.idx]);
+        }
+        else
+        {
+            if ( (xatp.idx >= nr_grant_frames(d->grant_table)) &&
+                 (xatp.idx < max_nr_grant_frames) )
+                gnttab_grow_table(d, xatp.idx + 1);
+
+            if ( xatp.idx < nr_grant_frames(d->grant_table) )
+                mfn = virt_to_mfn(d->grant_table->shared_raw[xatp.idx]);
         }
 
-        switch ( xatp.space )
+        spin_unlock(&d->grant_table->lock);
+        break;
+    case XENMAPSPACE_gmfn:
+    {
+        p2m_type_t p2mt;
+
+        xatp.idx = mfn_x(gfn_to_mfn_unshare(d, xatp.idx, &p2mt));
+        /* If the page is still shared, exit early */
+        if ( p2m_is_shared(p2mt) )
         {
-        case XENMAPSPACE_shared_info:
-            if ( xatp.idx == 0 )
-                mfn = virt_to_mfn(d->shared_info);
+            rcu_unlock_domain(d);
+            return -ENOMEM;
+        }
+        if ( !get_page_from_pagenr(xatp.idx, d) )
             break;
-        case XENMAPSPACE_grant_table:
-            spin_lock(&d->grant_table->lock);
+        mfn = xatp.idx;
+        page = mfn_to_page(mfn);
+        break;
+    }
+    default:
+        break;
+    }
 
-            if ( d->grant_table->gt_version == 0 )
-                d->grant_table->gt_version = 1;
+    if ( !paging_mode_translate(d) || (mfn == 0) )
+    {
+        if ( page )
+            put_page(page);
+        rcu_unlock_domain(d);
+        return -EINVAL;
+    }
 
-            if ( d->grant_table->gt_version == 2 &&
-                 (xatp.idx & XENMAPIDX_grant_table_status) )
-            {
-                xatp.idx &= ~XENMAPIDX_grant_table_status;
-                if ( xatp.idx < nr_status_frames(d->grant_table) )
-                    mfn = virt_to_mfn(d->grant_table->status[xatp.idx]);
-            }
-            else
-            {
-                if ( (xatp.idx >= nr_grant_frames(d->grant_table)) &&
-                     (xatp.idx < max_nr_grant_frames) )
-                    gnttab_grow_table(d, xatp.idx + 1);
+    domain_lock(d);
 
-                if ( xatp.idx < nr_grant_frames(d->grant_table) )
-                    mfn = virt_to_mfn(d->grant_table->shared_raw[xatp.idx]);
-            }
+    if ( page )
+        put_page(page);
 
-            spin_unlock(&d->grant_table->lock);
-            break;
-        case XENMAPSPACE_gmfn:
-        {
-            p2m_type_t p2mt;
+    /* Remove previously mapped page if it was present. */
+    prev_mfn = gmfn_to_mfn(d, xatp.gpfn);
+    if ( mfn_valid(prev_mfn) )
+    {
+        if ( is_xen_heap_mfn(prev_mfn) )
+            /* Xen heap frames are simply unhooked from this phys slot. */
+            guest_physmap_remove_page(d, xatp.gpfn, prev_mfn, 0);
+        else
+            /* Normal domain memory is freed, to avoid leaking memory. */
+            guest_remove_page(d, xatp.gpfn);
+    }
 
-            xatp.idx = mfn_x(gfn_to_mfn_unshare(d, xatp.idx, &p2mt));
-            /* If the page is still shared, exit early */
-            if ( p2m_is_shared(p2mt) )
-            {
-                rcu_unlock_domain(d);
-                return -ENOMEM;
-            }
-            if ( !get_page_from_pagenr(xatp.idx, d) )
-                break;
-            mfn = xatp.idx;
-            page = mfn_to_page(mfn);
-            break;
-        }
-        default:
-            break;
-        }
+    /* Unmap from old location, if any. */
+    gpfn = get_gpfn_from_mfn(mfn);
+    ASSERT( gpfn != SHARED_M2P_ENTRY );
+    if ( gpfn != INVALID_M2P_ENTRY )
+        guest_physmap_remove_page(d, gpfn, mfn, 0);
 
-        if ( !paging_mode_translate(d) || (mfn == 0) )
-        {
-            if ( page )
-                put_page(page);
-            rcu_unlock_domain(d);
-            return -EINVAL;
-        }
+    /* Map at new location. */
+    rc = guest_physmap_add_page(d, xatp.gpfn, mfn, 0);
 
-        domain_lock(d);
+    domain_unlock(d);
 
-        if ( page )
-            put_page(page);
+    return rc;
+}
 
-        /* Remove previously mapped page if it was present. */
-        prev_mfn = gmfn_to_mfn(d, xatp.gpfn);
-        if ( mfn_valid(prev_mfn) )
-        {
-            if ( is_xen_heap_mfn(prev_mfn) )
-                /* Xen heap frames are simply unhooked from this phys slot. */
-                guest_physmap_remove_page(d, xatp.gpfn, prev_mfn, 0);
-            else
-                /* Normal domain memory is freed, to avoid leaking memory. */
-                guest_remove_page(d, xatp.gpfn);
-        }
 
-        /* Unmap from old location, if any. */
-        gpfn = get_gpfn_from_mfn(mfn);
-        ASSERT( gpfn != SHARED_M2P_ENTRY );
-        if ( gpfn != INVALID_M2P_ENTRY )
-            guest_physmap_remove_page(d, gpfn, mfn, 0);
+long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
+{
+    int rc;
+
+    switch ( op )
+    {
+    case XENMEM_add_to_physmap:
+    {
+        struct xen_add_to_physmap xatp;
+        struct domain *d;
 
-        /* Map at new location. */
-        rc = guest_physmap_add_page(d, xatp.gpfn, mfn, 0);
+        if ( copy_from_guest(&xatp, arg, 1) )
+            return -EFAULT;
+
+        rc = rcu_lock_target_domain_by_id(xatp.domid, &d);
+        if ( rc != 0 )
+            return rc;
+
+        if ( xsm_add_to_physmap(current->domain, d) )
+        {
+            rcu_unlock_domain(d);
+            return -EPERM;
+        }
 
-        domain_unlock(d);
+        xenmem_add_to_physmap(d, xatp);
 
         rcu_unlock_domain(d);
 

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 471 bytes --]


XENMEM_add_to_physmap_gmfn_range is like XENMEM_add_to_physmap on
the gmfn space but the number of pages on which xen will iterate.

Use the 16 bits padding between .domid and .space in struct xen_add_to_physmap
to keep compatibility with older versions.

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/arch/x86/mm.c           |   22 +++++++++++++++++++++-
 xen/include/public/memory.h |    5 ++++-
 2 files changed, 25 insertions(+), 2 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0004-mm-New-XENMEM-XENMEM_add_to_physmap_gmfn_range.patch --]
[-- Type: text/x-patch; name="0004-mm-New-XENMEM-XENMEM_add_to_physmap_gmfn_range.patch", Size: 1862 bytes --]

diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index f75011e..cca64b8 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -4714,9 +4714,29 @@ long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
             return -EPERM;
         }
 
-        xenmem_add_to_physmap(d, xatp);
+        if ( xatp.space != XENMAPSPACE_gmfn_range )
+            xatp.size = 1;
+
+        for ( ; xatp.size > 0; xatp.size-- )
+        {
+            if ( hypercall_preempt_check() )
+            {
+                rc = -EAGAIN;
+                break;
+            }
+            xenmem_add_to_physmap(d, xatp);
+            xatp.idx++;
+            xatp.gpfn++;
+        }
 
         rcu_unlock_domain(d);
+        if ( rc == -EAGAIN )
+        {
+            if ( copy_to_guest(arg, &xatp, 1) )
+                return -EFAULT;
+            rc = hypercall_create_continuation(
+                    __HYPERVISOR_memory_op, "ih", op, arg);
+        }
 
         return rc;
     }
diff --git a/xen/include/public/memory.h b/xen/include/public/memory.h
index 08355e3..886d4c3 100644
--- a/xen/include/public/memory.h
+++ b/xen/include/public/memory.h
@@ -208,10 +208,14 @@ struct xen_add_to_physmap {
     /* Which domain to change the mapping for. */
     domid_t domid;
 
+    /* Number of pages to go through for gmfn_range */
+    uint16_t    size;
+
     /* Source mapping space. */
 #define XENMAPSPACE_shared_info 0 /* shared info page */
 #define XENMAPSPACE_grant_table 1 /* grant table page */
 #define XENMAPSPACE_gmfn        2 /* GMFN */
+#define XENMAPSPACE_gmfn_range  3 /* GMFN_range */
     unsigned int space;
 
 #define XENMAPIDX_grant_table_status 0x80000000
@@ -290,7 +294,6 @@ typedef struct xen_pod_target xen_pod_target_t;
  * The call never fails. 
  */
 #define XENMEM_get_sharing_freed_pages    18
-
 #endif /* __XEN_PUBLIC_MEMORY_H__ */
 
 /*

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 5/6] hvmloader: Change memory relocation loop when overlap with PCI hole.

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 715 bytes --]


Change the way we relocate the memory page if they overlap with pci hole.
Use new map space (XENMAPSPACE_gmfn_range) to move the loop into xen.

This code usually get triggered when a device is pass through to a guest
and the PCI hole has to be extended to have enough room to map the device BARs.
The PCI hole will starts lower and it might overlap with some RAM that has been
alocated for the guest. That usually happen if the guest has more than 4G of RAM.
We have to relocate those pages in high mem otherwise they won't be accessible.

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 tools/firmware/hvmloader/pci.c |   20 ++++++++++++++------
 1 files changed, 14 insertions(+), 6 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0005-hvmloader-Change-memory-relocation-loop-when-overlap.patch --]
[-- Type: text/x-patch; name="0005-hvmloader-Change-memory-relocation-loop-when-overlap.patch", Size: 1719 bytes --]

diff --git a/tools/firmware/hvmloader/pci.c b/tools/firmware/hvmloader/pci.c
index 29ec011..3bd6ac5 100644
--- a/tools/firmware/hvmloader/pci.c
+++ b/tools/firmware/hvmloader/pci.c
@@ -50,6 +50,7 @@ void pci_setup(void)
         uint32_t devfn, bar_reg, bar_sz;
     } *bars = (struct bars *)scratch_start;
     unsigned int i, nr_bars = 0;
+    unsigned long pci_mem_start_pg;
 
     /* Program PCI-ISA bridge with appropriate link routes. */
     isa_irq = 0;
@@ -185,17 +186,24 @@ void pci_setup(void)
             ((pci_mem_start << 1) != 0) )
         pci_mem_start <<= 1;
 
-    while ( (pci_mem_start >> PAGE_SHIFT) < hvm_info->low_mem_pgend )
+    /* Relocate RAM that overlaps (in 64K chunks) */
+    pci_mem_start_pg = (pci_mem_start >> PAGE_SHIFT);
+    while (pci_mem_start_pg < hvm_info->low_mem_pgend)
     {
         struct xen_add_to_physmap xatp;
-        if ( hvm_info->high_mem_pgend == 0 )
-            hvm_info->high_mem_pgend = 1ull << (32 - PAGE_SHIFT);
+        unsigned int size = hvm_info->low_mem_pgend - pci_mem_start_pg;
         xatp.domid = DOMID_SELF;
-        xatp.space = XENMAPSPACE_gmfn;
-        xatp.idx   = --hvm_info->low_mem_pgend;
-        xatp.gpfn  = hvm_info->high_mem_pgend++;
+        xatp.space = XENMAPSPACE_gmfn_range;
+        xatp.idx = pci_mem_start_pg;
+        xatp.gpfn = hvm_info->high_mem_pgend;
+        size = size > ((1 << 16) - 1) ? ((1 << 16) - 1) : size;
+        xatp.size = size;
+
         if ( hypercall_memory_op(XENMEM_add_to_physmap, &xatp) != 0 )
             BUG();
+        pci_mem_start_pg += size;
+        hvm_info->high_mem_pgend += size;
+        hvm_info->low_mem_pgend = pci_mem_start_pg;
     }
 
     mem_resource.base = pci_mem_start;

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 6/6] Introduce per cpu flag (iommu_dont_flush_iotlb) to avoid unnecessary iotlb flush

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 454 bytes --]


Add cpu flag that will be checked by the iommu low level code
to skip iotlb flushes. iommu_iotlb_flush shall be called explicitly.

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/arch/x86/mm.c                   |   15 ++++++++++++++-
 xen/drivers/passthrough/iommu.c     |    5 +++++
 xen/drivers/passthrough/vtd/iommu.c |    6 ++++--
 xen/include/xen/iommu.h             |    2 ++
 4 files changed, 25 insertions(+), 3 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0006-Introduce-per-cpu-flag-iommu_dont_flush_iotlb-to-avo.patch --]
[-- Type: text/x-patch; name="0006-Introduce-per-cpu-flag-iommu_dont_flush_iotlb-to-avo.patch", Size: 3674 bytes --]

diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index cca64b8..7aece56 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -4698,7 +4698,7 @@ long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
     {
     case XENMEM_add_to_physmap:
     {
-        struct xen_add_to_physmap xatp;
+        struct xen_add_to_physmap xatp, start_xatp;
         struct domain *d;
 
         if ( copy_from_guest(&xatp, arg, 1) )
@@ -4716,7 +4716,13 @@ long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
 
         if ( xatp.space != XENMAPSPACE_gmfn_range )
             xatp.size = 1;
+        else
+        {
+            if ( need_iommu(d) )
+                this_cpu(iommu_dont_flush_iotlb) = 1;
+        }
 
+        start_xatp = xatp;
         for ( ; xatp.size > 0; xatp.size-- )
         {
             if ( hypercall_preempt_check() )
@@ -4729,6 +4735,13 @@ long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
             xatp.gpfn++;
         }
 
+        if ( xatp.space == XENMAPSPACE_gmfn_range && need_iommu(d) )
+        {
+            this_cpu(iommu_dont_flush_iotlb) = 0;
+            iommu_iotlb_flush(d, start_xatp.idx, start_xatp.size - xatp.size);
+            iommu_iotlb_flush(d, start_xatp.gpfn, start_xatp.size - xatp.size);
+        }
+
         rcu_unlock_domain(d);
         if ( rc == -EAGAIN )
         {
diff --git a/xen/drivers/passthrough/iommu.c b/xen/drivers/passthrough/iommu.c
index 9c62861..a06c94a 100644
--- a/xen/drivers/passthrough/iommu.c
+++ b/xen/drivers/passthrough/iommu.c
@@ -52,6 +52,8 @@ bool_t __read_mostly iommu_hap_pt_share = 1;
 bool_t __read_mostly iommu_debug;
 bool_t __read_mostly amd_iommu_perdev_intremap;
 
+DEFINE_PER_CPU(bool_t, iommu_dont_flush_iotlb);
+
 static void __init parse_iommu_param(char *s)
 {
     char *ss;
@@ -227,6 +229,7 @@ static int iommu_populate_page_table(struct domain *d)
 
     spin_lock(&d->page_alloc_lock);
 
+    this_cpu(iommu_dont_flush_iotlb) = 1;
     page_list_for_each ( page, &d->page_list )
     {
         if ( is_hvm_domain(d) ||
@@ -244,6 +247,8 @@ static int iommu_populate_page_table(struct domain *d)
             }
         }
     }
+    this_cpu(iommu_dont_flush_iotlb) = 0;
+    iommu_iotlb_flush_all(d);
     spin_unlock(&d->page_alloc_lock);
     return 0;
 }
diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 7ec9541..a3dd018 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -660,7 +660,8 @@ static void dma_pte_clear_one(struct domain *domain, u64 addr)
     spin_unlock(&hd->mapping_lock);
     iommu_flush_cache_entry(pte, sizeof(struct dma_pte));
 
-    __intel_iommu_iotlb_flush(domain, addr >> PAGE_SHIFT_4K , 0, 1);
+    if ( !this_cpu(iommu_dont_flush_iotlb) )
+        __intel_iommu_iotlb_flush(domain, addr >> PAGE_SHIFT_4K , 0, 1);
 
     unmap_vtd_domain_page(page);
 
@@ -1753,7 +1754,8 @@ static int intel_iommu_map_page(
     spin_unlock(&hd->mapping_lock);
     unmap_vtd_domain_page(page);
 
-    __intel_iommu_iotlb_flush(d, gfn, dma_pte_present(old), 1);
+    if ( !this_cpu(iommu_dont_flush_iotlb) )
+        __intel_iommu_iotlb_flush(d, gfn, dma_pte_present(old), 1);
 
     return 0;
 }
diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
index a1034df..c757a78 100644
--- a/xen/include/xen/iommu.h
+++ b/xen/include/xen/iommu.h
@@ -160,4 +160,6 @@ int iommu_do_domctl(struct xen_domctl *, XEN_GUEST_HANDLE(xen_domctl_t));
 void iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count);
 void iommu_iotlb_flush_all(struct domain *d);
 
+DECLARE_PER_CPU(bool_t, iommu_dont_flush_iotlb);
+
 #endif /* _IOMMU_H_ */

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Jan Beulich]

>>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:

> XENMEM_add_to_physmap_gmfn_range is like XENMEM_add_to_physmap on
> the gmfn space but the number of pages on which xen will iterate.
> 
> Use the 16 bits padding between .domid and .space in struct 
> xen_add_to_physmap
> to keep compatibility with older versions.
> 
> Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> ---
>  xen/arch/x86/mm.c           |   22 +++++++++++++++++++++-
>  xen/include/public/memory.h |    5 ++++-
>  2 files changed, 25 insertions(+), 2 deletions(-)

Please remove the stray last hunk of the patch (the newline there
should stay).

Jan

Re: [PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Jean Guyader]

On 07/11 04:45, Jan Beulich wrote:
> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> 
> > XENMEM_add_to_physmap_gmfn_range is like XENMEM_add_to_physmap on
> > the gmfn space but the number of pages on which xen will iterate.
> > 
> > Use the 16 bits padding between .domid and .space in struct 
> > xen_add_to_physmap
> > to keep compatibility with older versions.
> > 
> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> > ---
> >  xen/arch/x86/mm.c           |   22 +++++++++++++++++++++-
> >  xen/include/public/memory.h |    5 ++++-
> >  2 files changed, 25 insertions(+), 2 deletions(-)
> 
> Please remove the stray last hunk of the patch (the newline there
> should stay).
> 

Yes, it shouldn't be there. I've updated the patch.

Thanks,
Jean

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jan Beulich]

>>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:

> Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> ---
>  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
>  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
>  xen/include/xen/iommu.h             |    5 +++++
>  3 files changed, 37 insertions(+), 0 deletions(-)

In iommu_iotlb_flush() you check whether the to-be-called function
pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
actually think the second behavior is the correct one, but that
implies that you need to also implement respective AMD IOMMU
functions.

Jan

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jean Guyader]

On 07/11 04:42, Jan Beulich wrote:
> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> 
> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> > ---
> >  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
> >  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
> >  xen/include/xen/iommu.h             |    5 +++++
> >  3 files changed, 37 insertions(+), 0 deletions(-)
> 
> In iommu_iotlb_flush() you check whether the to-be-called function
> pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
> actually think the second behavior is the correct one, but that
> implies that you need to also implement respective AMD IOMMU
> functions.
> 

Yes, It's an error on my part. I've updated the patch to check
for the present of iotlb_flush_all before I call it now.

Thanks,
Jean

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jan Beulich]

>>> On 07.11.11 at 18:06, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> On 07/11 04:42, Jan Beulich wrote:
>> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
>> 
>> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
>> > ---
>> >  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
>> >  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
>> >  xen/include/xen/iommu.h             |    5 +++++
>> >  3 files changed, 37 insertions(+), 0 deletions(-)
>> 
>> In iommu_iotlb_flush() you check whether the to-be-called function
>> pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
>> actually think the second behavior is the correct one, but that
>> implies that you need to also implement respective AMD IOMMU
>> functions.
>> 
> 
> Yes, It's an error on my part. I've updated the patch to check
> for the present of iotlb_flush_all before I call it now.

But as said, I don't think this is the right solution: How can it be
correct on non-Intel hardware to have these functions simply do
nothing?

Jan

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jean Guyader]

On 08/11 08:24, Jan Beulich wrote:
> >>> On 07.11.11 at 18:06, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> > On 07/11 04:42, Jan Beulich wrote:
> >> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> >> 
> >> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> >> > ---
> >> >  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
> >> >  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
> >> >  xen/include/xen/iommu.h             |    5 +++++
> >> >  3 files changed, 37 insertions(+), 0 deletions(-)
> >> 
> >> In iommu_iotlb_flush() you check whether the to-be-called function
> >> pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
> >> actually think the second behavior is the correct one, but that
> >> implies that you need to also implement respective AMD IOMMU
> >> functions.
> >> 
> > 
> > Yes, It's an error on my part. I've updated the patch to check
> > for the present of iotlb_flush_all before I call it now.
> 
> But as said, I don't think this is the right solution: How can it be
> correct on non-Intel hardware to have these functions simply do
> nothing?
> 

These functions are only here to counter balance the dont_flush_iotlb_flag.
If they don't acknowlage this flag doing nothing is the right thing to do.

That said I can probably implement a naive version for the AMD iommu
that will use amd_iommu_flush_pages once or in a loop.

Obviously I will remove the check for the non existing callback if I do
that as it become mandatory.

Jean

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jan Beulich]

>>> On 08.11.11 at 11:25, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> On 08/11 08:24, Jan Beulich wrote:
>> >>> On 07.11.11 at 18:06, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
>> > On 07/11 04:42, Jan Beulich wrote:
>> >> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
>> >> 
>> >> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
>> >> > ---
>> >> >  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
>> >> >  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
>> >> >  xen/include/xen/iommu.h             |    5 +++++
>> >> >  3 files changed, 37 insertions(+), 0 deletions(-)
>> >> 
>> >> In iommu_iotlb_flush() you check whether the to-be-called function
>> >> pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
>> >> actually think the second behavior is the correct one, but that
>> >> implies that you need to also implement respective AMD IOMMU
>> >> functions.
>> >> 
>> > 
>> > Yes, It's an error on my part. I've updated the patch to check
>> > for the present of iotlb_flush_all before I call it now.
>> 
>> But as said, I don't think this is the right solution: How can it be
>> correct on non-Intel hardware to have these functions simply do
>> nothing?
>> 
> 
> These functions are only here to counter balance the dont_flush_iotlb_flag.
> If they don't acknowlage this flag doing nothing is the right thing to do.

Ah, okay, I overlooked this aspect. Might be worth a comment in
the header file.

Jan

> That said I can probably implement a naive version for the AMD iommu
> that will use amd_iommu_flush_pages once or in a loop.
> 
> Obviously I will remove the check for the non existing callback if I do
> that as it become mandatory.
> 
> Jean

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Wei Wang2]

On Tuesday 08 November 2011 11:25:15 Jean Guyader wrote:
> On 08/11 08:24, Jan Beulich wrote:
> > >>> On 07.11.11 at 18:06, Jean Guyader <jean.guyader@eu.citrix.com>
> > >>> wrote:
> > >
> > > On 07/11 04:42, Jan Beulich wrote:
> > >> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com>
> > >> >>> wrote:
> > >> >
> > >> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> > >> > ---
> > >> >  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
> > >> >  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
> > >> >  xen/include/xen/iommu.h             |    5 +++++
> > >> >  3 files changed, 37 insertions(+), 0 deletions(-)
> > >>
> > >> In iommu_iotlb_flush() you check whether the to-be-called function
> > >> pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
> > >> actually think the second behavior is the correct one, but that
> > >> implies that you need to also implement respective AMD IOMMU
> > >> functions.
> > >
> > > Yes, It's an error on my part. I've updated the patch to check
> > > for the present of iotlb_flush_all before I call it now.
> >
> > But as said, I don't think this is the right solution: How can it be
> > correct on non-Intel hardware to have these functions simply do
> > nothing?
>
> These functions are only here to counter balance the dont_flush_iotlb_flag.
> If they don't acknowlage this flag doing nothing is the right thing to do.
>
> That said I can probably implement a naive version for the AMD iommu
> that will use amd_iommu_flush_pages once or in a loop.

There are some pending patches that enable ats / iotlb flush for amd iommu. 
Then, you might want to use functions: amd_iommu_flush_iotlb and 
amd_iommu_flush_all_iotlbs for this.
Thanks,
Wei

> Obviously I will remove the check for the non existing callback if I do
> that as it become mandatory.
>
> Jean
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Re: [PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jean Guyader]

On 08/11 10:39, Wei Wang2 wrote:
> On Tuesday 08 November 2011 11:25:15 Jean Guyader wrote:
> > On 08/11 08:24, Jan Beulich wrote:
> > > >>> On 07.11.11 at 18:06, Jean Guyader <jean.guyader@eu.citrix.com>
> > > >>> wrote:
> > > >
> > > > On 07/11 04:42, Jan Beulich wrote:
> > > >> >>> On 07.11.11 at 16:16, Jean Guyader <jean.guyader@eu.citrix.com>
> > > >> >>> wrote:
> > > >> >
> > > >> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> > > >> > ---
> > > >> >  xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
> > > >> >  xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
> > > >> >  xen/include/xen/iommu.h             |    5 +++++
> > > >> >  3 files changed, 37 insertions(+), 0 deletions(-)
> > > >>
> > > >> In iommu_iotlb_flush() you check whether the to-be-called function
> > > >> pointer is NULL, whereas in iommu_iotlb_flush_all() you don't. I
> > > >> actually think the second behavior is the correct one, but that
> > > >> implies that you need to also implement respective AMD IOMMU
> > > >> functions.
> > > >
> > > > Yes, It's an error on my part. I've updated the patch to check
> > > > for the present of iotlb_flush_all before I call it now.
> > >
> > > But as said, I don't think this is the right solution: How can it be
> > > correct on non-Intel hardware to have these functions simply do
> > > nothing?
> >
> > These functions are only here to counter balance the dont_flush_iotlb_flag.
> > If they don't acknowlage this flag doing nothing is the right thing to do.
> >
> > That said I can probably implement a naive version for the AMD iommu
> > that will use amd_iommu_flush_pages once or in a loop.
> 
> There are some pending patches that enable ats / iotlb flush for amd iommu. 
> Then, you might want to use functions: amd_iommu_flush_iotlb and 
> amd_iommu_flush_all_iotlbs for this.
> Thanks,
> Wei
> 

Ok, I'll do that in the next version.

Thanks,
Jean

IOMMU, vtd and iotlb flush rework (v3)

[Jean Guyader]

In one of my previous email I detailed a bug I was seeing when passing
through a Intel GPU on a guest that has more that 4G or RAM.

Allen suggested that I go for the Plan B but after a discussion with Tim
we agreed that Plan B was way to disruptive in term of code change.

This patch series implements Plan A.

http://xen.1045712.n5.nabble.com/VTD-Intel-iommu-IOTLB-flush-really-slow-td4952866.html

Changes between v2 and v3:
        - Check for the presence iotlb_flush_all callback before calling it.

Changes between v1 and v2:
       - Move size in struct xen_add_to_physmap in padding between .domid and .space.
       - Store iommu_dont_flush per cpu
       - Change the code in hvmloader to relocate by batch of 64K, .size is now 16 bits.

Jean Guyader (6):
      vtd: Refactor iotlb flush code
      iommu: Introduce iommu_flush and iommu_flush_all.
      add_to_physmap: Move the code for XENMEM_add_to_physmap.
      mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range
      hvmloader: Change memory relocation loop when overlap with PCI hole.
      Introduce per cpu flag (iommu_dont_flush_iotlb) to avoid unnecessary iotlb flush

 tools/firmware/hvmloader/pci.c      |   20 +++-
 xen/arch/x86/mm.c                   |  203 +++++++++++++++++++++--------------
 xen/drivers/passthrough/iommu.c     |   25 +++++
 xen/drivers/passthrough/vtd/iommu.c |  100 ++++++++++--------
 xen/include/public/memory.h         |    4 +
 xen/include/xen/iommu.h             |    7 ++
 6 files changed, 230 insertions(+), 129 deletions(-)

[PATCH 1/6] vtd: Refactor iotlb flush code

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 279 bytes --]


Factorize the iotlb flush code from map_page and unmap_page into
it's own function.

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/drivers/passthrough/vtd/iommu.c |   86 +++++++++++++++++-----------------
 1 files changed, 43 insertions(+), 43 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-vtd-Refactor-iotlb-flush-code.patch --]
[-- Type: text/x-patch; name="0001-vtd-Refactor-iotlb-flush-code.patch", Size: 4122 bytes --]

diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 7717ab4..5a5b6be 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -578,16 +578,53 @@ static void iommu_flush_all(void)
     }
 }
 
+static void __intel_iommu_iotlb_flush(struct domain *d, unsigned long gfn,
+        int dma_old_pte_present, unsigned int page_count)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(d);
+    struct acpi_drhd_unit *drhd;
+    struct iommu *iommu;
+    int flush_dev_iotlb;
+    int iommu_domid;
+
+    /*
+     * No need pcideves_lock here because we have flush
+     * when assign/deassign device
+     */
+    for_each_drhd_unit ( drhd )
+    {
+        iommu = drhd->iommu;
+
+        if ( !test_bit(iommu->index, &hd->iommu_bitmap) )
+            continue;
+
+        flush_dev_iotlb = find_ats_dev_drhd(iommu) ? 1 : 0;
+        iommu_domid= domain_iommu_domid(d, iommu);
+        if ( iommu_domid == -1 )
+            continue;
+
+        if ( page_count > 1 || gfn == -1 )
+        {
+            if ( iommu_flush_iotlb_dsi(iommu, iommu_domid,
+                        0, flush_dev_iotlb) )
+                iommu_flush_write_buffer(iommu);
+        }
+        else
+        {
+            if ( iommu_flush_iotlb_psi(iommu, iommu_domid,
+                        (paddr_t)gfn << PAGE_SHIFT_4K, 0,
+                        !dma_old_pte_present, flush_dev_iotlb) )
+                iommu_flush_write_buffer(iommu);
+        }
+    }
+}
+
 /* clear one page's page table */
 static void dma_pte_clear_one(struct domain *domain, u64 addr)
 {
     struct hvm_iommu *hd = domain_hvm_iommu(domain);
-    struct acpi_drhd_unit *drhd;
-    struct iommu *iommu;
     struct dma_pte *page = NULL, *pte = NULL;
     u64 pg_maddr;
-    int flush_dev_iotlb;
-    int iommu_domid;
     struct mapped_rmrr *mrmrr;
 
     spin_lock(&hd->mapping_lock);
@@ -613,21 +650,7 @@ static void dma_pte_clear_one(struct domain *domain, u64 addr)
     spin_unlock(&hd->mapping_lock);
     iommu_flush_cache_entry(pte, sizeof(struct dma_pte));
 
-    /* No need pcidevs_lock here since do that on assign/deassign device*/
-    for_each_drhd_unit ( drhd )
-    {
-        iommu = drhd->iommu;
-        if ( test_bit(iommu->index, &hd->iommu_bitmap) )
-        {
-            flush_dev_iotlb = find_ats_dev_drhd(iommu) ? 1 : 0;
-            iommu_domid= domain_iommu_domid(domain, iommu);
-            if ( iommu_domid == -1 )
-                continue;
-            if ( iommu_flush_iotlb_psi(iommu, iommu_domid, addr,
-                                       0, 0, flush_dev_iotlb) )
-                iommu_flush_write_buffer(iommu);
-        }
-    }
+    __intel_iommu_iotlb_flush(domain, addr >> PAGE_SHIFT_4K , 0, 1);
 
     unmap_vtd_domain_page(page);
 
@@ -1677,12 +1700,8 @@ static int intel_iommu_map_page(
     unsigned int flags)
 {
     struct hvm_iommu *hd = domain_hvm_iommu(d);
-    struct acpi_drhd_unit *drhd;
-    struct iommu *iommu;
     struct dma_pte *page = NULL, *pte = NULL, old, new = { 0 };
     u64 pg_maddr;
-    int flush_dev_iotlb;
-    int iommu_domid;
 
     /* Do nothing if VT-d shares EPT page table */
     if ( iommu_use_hap_pt(d) )
@@ -1724,26 +1743,7 @@ static int intel_iommu_map_page(
     spin_unlock(&hd->mapping_lock);
     unmap_vtd_domain_page(page);
 
-    /*
-     * No need pcideves_lock here because we have flush
-     * when assign/deassign device
-     */
-    for_each_drhd_unit ( drhd )
-    {
-        iommu = drhd->iommu;
-
-        if ( !test_bit(iommu->index, &hd->iommu_bitmap) )
-            continue;
-
-        flush_dev_iotlb = find_ats_dev_drhd(iommu) ? 1 : 0;
-        iommu_domid= domain_iommu_domid(d, iommu);
-        if ( iommu_domid == -1 )
-            continue;
-        if ( iommu_flush_iotlb_psi(iommu, iommu_domid,
-                                   (paddr_t)gfn << PAGE_SHIFT_4K, 0,
-                                   !dma_pte_present(old), flush_dev_iotlb) )
-            iommu_flush_write_buffer(iommu);
-    }
+    __intel_iommu_iotlb_flush(d, gfn, dma_pte_present(old), 1);
 
     return 0;
 }

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 2/6] iommu: Introduce iommu_flush and iommu_flush_all.

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 286 bytes --]


Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/drivers/passthrough/iommu.c     |   20 ++++++++++++++++++++
 xen/drivers/passthrough/vtd/iommu.c |   12 ++++++++++++
 xen/include/xen/iommu.h             |    5 +++++
 3 files changed, 37 insertions(+), 0 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0002-iommu-Introduce-iommu_flush-and-iommu_flush_all.patch --]
[-- Type: text/x-patch; name="0002-iommu-Introduce-iommu_flush-and-iommu_flush_all.patch", Size: 2780 bytes --]

diff --git a/xen/drivers/passthrough/iommu.c b/xen/drivers/passthrough/iommu.c
index cd6174d..ca7b37b 100644
--- a/xen/drivers/passthrough/iommu.c
+++ b/xen/drivers/passthrough/iommu.c
@@ -301,6 +301,26 @@ int iommu_unmap_page(struct domain *d, unsigned long gfn)
     return hd->platform_ops->unmap_page(d, gfn);
 }
 
+void iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(d);
+
+    if ( !iommu_enabled || !hd->platform_ops || !hd->platform_ops->iotlb_flush )
+        return;
+
+    hd->platform_ops->iotlb_flush(d, gfn, page_count);
+}
+
+void iommu_iotlb_flush_all(struct domain *d)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(d);
+
+    if ( !iommu_enabled || !hd->platform_ops || !hd->platform_ops->iotlb_flush_all )
+        return;
+
+    hd->platform_ops->iotlb_flush_all(d);
+}
+
 /* caller should hold the pcidevs_lock */
 int deassign_device(struct domain *d, u16 seg, u8 bus, u8 devfn)
 {
diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 5a5b6be..7ec9541 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -619,6 +619,16 @@ static void __intel_iommu_iotlb_flush(struct domain *d, unsigned long gfn,
     }
 }
 
+static void intel_iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count)
+{
+    __intel_iommu_iotlb_flush(d, gfn, 1, page_count);
+}
+
+static void intel_iommu_iotlb_flush_all(struct domain *d)
+{
+    __intel_iommu_iotlb_flush(d, 0, 0, 0);
+}
+
 /* clear one page's page table */
 static void dma_pte_clear_one(struct domain *domain, u64 addr)
 {
@@ -2329,6 +2339,8 @@ const struct iommu_ops intel_iommu_ops = {
     .resume = vtd_resume,
     .share_p2m = iommu_set_pgd,
     .crash_shutdown = vtd_crash_shutdown,
+    .iotlb_flush = intel_iommu_iotlb_flush,
+    .iotlb_flush_all = intel_iommu_iotlb_flush_all,
 };
 
 /*
diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
index 837e60d..a1034df 100644
--- a/xen/include/xen/iommu.h
+++ b/xen/include/xen/iommu.h
@@ -139,6 +139,8 @@ struct iommu_ops {
     void (*resume)(void);
     void (*share_p2m)(struct domain *d);
     void (*crash_shutdown)(void);
+    void (*iotlb_flush)(struct domain *d, unsigned long gfn, unsigned int page_count);
+    void (*iotlb_flush_all)(struct domain *d);
 };
 
 void iommu_update_ire_from_apic(unsigned int apic, unsigned int reg, unsigned int value);
@@ -155,4 +157,7 @@ void iommu_share_p2m_table(struct domain *d);
 
 int iommu_do_domctl(struct xen_domctl *, XEN_GUEST_HANDLE(xen_domctl_t));
 
+void iommu_iotlb_flush(struct domain *d, unsigned long gfn, unsigned int page_count);
+void iommu_iotlb_flush_all(struct domain *d);
+
 #endif /* _IOMMU_H_ */

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 3/6] add_to_physmap: Move the code for XENMEM_add_to_physmap.

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 293 bytes --]


Move the code for the XENMEM_add_to_physmap case into it's own
function (xenmem_add_to_physmap).

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/arch/x86/mm.c |  188 ++++++++++++++++++++++++++++-------------------------
 1 files changed, 99 insertions(+), 89 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0003-add_to_physmap-Move-the-code-for-XENMEM_add_to_physm.patch --]
[-- Type: text/x-patch; name="0003-add_to_physmap-Move-the-code-for-XENMEM_add_to_physm.patch", Size: 7065 bytes --]

diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index acc1f34..f75011e 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -4592,119 +4592,129 @@ static int handle_iomem_range(unsigned long s, unsigned long e, void *p)
     return 0;
 }
 
-long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
+static int xenmem_add_to_physmap(struct domain *d, struct xen_add_to_physmap xatp)
 {
     struct page_info *page = NULL;
+    unsigned long prev_mfn, mfn = 0, gpfn;
     int rc;
 
-    switch ( op )
-    {
-    case XENMEM_add_to_physmap:
+    switch ( xatp.space )
     {
-        struct xen_add_to_physmap xatp;
-        unsigned long prev_mfn, mfn = 0, gpfn;
-        struct domain *d;
-
-        if ( copy_from_guest(&xatp, arg, 1) )
-            return -EFAULT;
+    case XENMAPSPACE_shared_info:
+        if ( xatp.idx == 0 )
+            mfn = virt_to_mfn(d->shared_info);
+        break;
+    case XENMAPSPACE_grant_table:
+        spin_lock(&d->grant_table->lock);
 
-        rc = rcu_lock_target_domain_by_id(xatp.domid, &d);
-        if ( rc != 0 )
-            return rc;
+        if ( d->grant_table->gt_version == 0 )
+            d->grant_table->gt_version = 1;
 
-        if ( xsm_add_to_physmap(current->domain, d) )
+        if ( d->grant_table->gt_version == 2 &&
+             (xatp.idx & XENMAPIDX_grant_table_status) )
         {
-            rcu_unlock_domain(d);
-            return -EPERM;
+            xatp.idx &= ~XENMAPIDX_grant_table_status;
+            if ( xatp.idx < nr_status_frames(d->grant_table) )
+                mfn = virt_to_mfn(d->grant_table->status[xatp.idx]);
+        }
+        else
+        {
+            if ( (xatp.idx >= nr_grant_frames(d->grant_table)) &&
+                 (xatp.idx < max_nr_grant_frames) )
+                gnttab_grow_table(d, xatp.idx + 1);
+
+            if ( xatp.idx < nr_grant_frames(d->grant_table) )
+                mfn = virt_to_mfn(d->grant_table->shared_raw[xatp.idx]);
         }
 
-        switch ( xatp.space )
+        spin_unlock(&d->grant_table->lock);
+        break;
+    case XENMAPSPACE_gmfn:
+    {
+        p2m_type_t p2mt;
+
+        xatp.idx = mfn_x(gfn_to_mfn_unshare(d, xatp.idx, &p2mt));
+        /* If the page is still shared, exit early */
+        if ( p2m_is_shared(p2mt) )
         {
-        case XENMAPSPACE_shared_info:
-            if ( xatp.idx == 0 )
-                mfn = virt_to_mfn(d->shared_info);
+            rcu_unlock_domain(d);
+            return -ENOMEM;
+        }
+        if ( !get_page_from_pagenr(xatp.idx, d) )
             break;
-        case XENMAPSPACE_grant_table:
-            spin_lock(&d->grant_table->lock);
+        mfn = xatp.idx;
+        page = mfn_to_page(mfn);
+        break;
+    }
+    default:
+        break;
+    }
 
-            if ( d->grant_table->gt_version == 0 )
-                d->grant_table->gt_version = 1;
+    if ( !paging_mode_translate(d) || (mfn == 0) )
+    {
+        if ( page )
+            put_page(page);
+        rcu_unlock_domain(d);
+        return -EINVAL;
+    }
 
-            if ( d->grant_table->gt_version == 2 &&
-                 (xatp.idx & XENMAPIDX_grant_table_status) )
-            {
-                xatp.idx &= ~XENMAPIDX_grant_table_status;
-                if ( xatp.idx < nr_status_frames(d->grant_table) )
-                    mfn = virt_to_mfn(d->grant_table->status[xatp.idx]);
-            }
-            else
-            {
-                if ( (xatp.idx >= nr_grant_frames(d->grant_table)) &&
-                     (xatp.idx < max_nr_grant_frames) )
-                    gnttab_grow_table(d, xatp.idx + 1);
+    domain_lock(d);
 
-                if ( xatp.idx < nr_grant_frames(d->grant_table) )
-                    mfn = virt_to_mfn(d->grant_table->shared_raw[xatp.idx]);
-            }
+    if ( page )
+        put_page(page);
 
-            spin_unlock(&d->grant_table->lock);
-            break;
-        case XENMAPSPACE_gmfn:
-        {
-            p2m_type_t p2mt;
+    /* Remove previously mapped page if it was present. */
+    prev_mfn = gmfn_to_mfn(d, xatp.gpfn);
+    if ( mfn_valid(prev_mfn) )
+    {
+        if ( is_xen_heap_mfn(prev_mfn) )
+            /* Xen heap frames are simply unhooked from this phys slot. */
+            guest_physmap_remove_page(d, xatp.gpfn, prev_mfn, 0);
+        else
+            /* Normal domain memory is freed, to avoid leaking memory. */
+            guest_remove_page(d, xatp.gpfn);
+    }
 
-            xatp.idx = mfn_x(gfn_to_mfn_unshare(d, xatp.idx, &p2mt));
-            /* If the page is still shared, exit early */
-            if ( p2m_is_shared(p2mt) )
-            {
-                rcu_unlock_domain(d);
-                return -ENOMEM;
-            }
-            if ( !get_page_from_pagenr(xatp.idx, d) )
-                break;
-            mfn = xatp.idx;
-            page = mfn_to_page(mfn);
-            break;
-        }
-        default:
-            break;
-        }
+    /* Unmap from old location, if any. */
+    gpfn = get_gpfn_from_mfn(mfn);
+    ASSERT( gpfn != SHARED_M2P_ENTRY );
+    if ( gpfn != INVALID_M2P_ENTRY )
+        guest_physmap_remove_page(d, gpfn, mfn, 0);
 
-        if ( !paging_mode_translate(d) || (mfn == 0) )
-        {
-            if ( page )
-                put_page(page);
-            rcu_unlock_domain(d);
-            return -EINVAL;
-        }
+    /* Map at new location. */
+    rc = guest_physmap_add_page(d, xatp.gpfn, mfn, 0);
 
-        domain_lock(d);
+    domain_unlock(d);
 
-        if ( page )
-            put_page(page);
+    return rc;
+}
 
-        /* Remove previously mapped page if it was present. */
-        prev_mfn = gmfn_to_mfn(d, xatp.gpfn);
-        if ( mfn_valid(prev_mfn) )
-        {
-            if ( is_xen_heap_mfn(prev_mfn) )
-                /* Xen heap frames are simply unhooked from this phys slot. */
-                guest_physmap_remove_page(d, xatp.gpfn, prev_mfn, 0);
-            else
-                /* Normal domain memory is freed, to avoid leaking memory. */
-                guest_remove_page(d, xatp.gpfn);
-        }
 
-        /* Unmap from old location, if any. */
-        gpfn = get_gpfn_from_mfn(mfn);
-        ASSERT( gpfn != SHARED_M2P_ENTRY );
-        if ( gpfn != INVALID_M2P_ENTRY )
-            guest_physmap_remove_page(d, gpfn, mfn, 0);
+long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
+{
+    int rc;
+
+    switch ( op )
+    {
+    case XENMEM_add_to_physmap:
+    {
+        struct xen_add_to_physmap xatp;
+        struct domain *d;
 
-        /* Map at new location. */
-        rc = guest_physmap_add_page(d, xatp.gpfn, mfn, 0);
+        if ( copy_from_guest(&xatp, arg, 1) )
+            return -EFAULT;
+
+        rc = rcu_lock_target_domain_by_id(xatp.domid, &d);
+        if ( rc != 0 )
+            return rc;
+
+        if ( xsm_add_to_physmap(current->domain, d) )
+        {
+            rcu_unlock_domain(d);
+            return -EPERM;
+        }
 
-        domain_unlock(d);
+        xenmem_add_to_physmap(d, xatp);
 
         rcu_unlock_domain(d);
 

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

[PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Jean Guyader]

[-- Attachment #1: Type: text/plain, Size: 470 bytes --]


XENMEM_add_to_physmap_gmfn_range is like XENMEM_add_to_physmap on
the gmfn space but the number of pages on which xen will iterate.

Use the 16 bits padding between .domid and .space in struct xen_add_to_physmap
to keep compatibility with older versions.

Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
---
 xen/arch/x86/mm.c           |   22 +++++++++++++++++++++-
 xen/include/public/memory.h |    4 ++++
 2 files changed, 25 insertions(+), 1 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0004-mm-New-XENMEM-XENMEM_add_to_physmap_gmfn_range.patch --]
[-- Type: text/x-patch; name="0004-mm-New-XENMEM-XENMEM_add_to_physmap_gmfn_range.patch", Size: 1670 bytes --]

diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index f75011e..cca64b8 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -4714,9 +4714,29 @@ long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
             return -EPERM;
         }
 
-        xenmem_add_to_physmap(d, xatp);
+        if ( xatp.space != XENMAPSPACE_gmfn_range )
+            xatp.size = 1;
+
+        for ( ; xatp.size > 0; xatp.size-- )
+        {
+            if ( hypercall_preempt_check() )
+            {
+                rc = -EAGAIN;
+                break;
+            }
+            xenmem_add_to_physmap(d, xatp);
+            xatp.idx++;
+            xatp.gpfn++;
+        }
 
         rcu_unlock_domain(d);
+        if ( rc == -EAGAIN )
+        {
+            if ( copy_to_guest(arg, &xatp, 1) )
+                return -EFAULT;
+            rc = hypercall_create_continuation(
+                    __HYPERVISOR_memory_op, "ih", op, arg);
+        }
 
         return rc;
     }
diff --git a/xen/include/public/memory.h b/xen/include/public/memory.h
index 08355e3..8bc8272 100644
--- a/xen/include/public/memory.h
+++ b/xen/include/public/memory.h
@@ -208,10 +208,14 @@ struct xen_add_to_physmap {
     /* Which domain to change the mapping for. */
     domid_t domid;
 
+    /* Number of pages to go through for gmfn_range */
+    uint16_t    size;
+
     /* Source mapping space. */
 #define XENMAPSPACE_shared_info 0 /* shared info page */
 #define XENMAPSPACE_grant_table 1 /* grant table page */
 #define XENMAPSPACE_gmfn        2 /* GMFN */
+#define XENMAPSPACE_gmfn_range  3 /* GMFN_range */
     unsigned int space;
 
 #define XENMAPIDX_grant_table_status 0x80000000

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Jan Beulich]

>>> On 07.11.11 at 19:25, Jean Guyader <jean.guyader@eu.citrix.com> wrote:

Sorry, noticed this only now, but neither title nor description of this
are in sync with the actual patch.

Jan

> XENMEM_add_to_physmap_gmfn_range is like XENMEM_add_to_physmap on
> the gmfn space but the number of pages on which xen will iterate.
> 
> Use the 16 bits padding between .domid and .space in struct 
> xen_add_to_physmap
> to keep compatibility with older versions.
> 
> Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> ---
>  xen/arch/x86/mm.c           |   22 +++++++++++++++++++++-
>  xen/include/public/memory.h |    4 ++++
>  2 files changed, 25 insertions(+), 1 deletions(-)

Re: [PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Tim Deegan]

At 02:20 -0700 on 08 Nov (1320718835), Jan Beulich wrote:
> >>> On 07.11.11 at 19:25, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> 
> Sorry, noticed this only now, but neither title nor description of this
> are in sync with the actual patch.


Indeed; they should be updated.  But otherwise:
Acked-by: Tim Deegan <tim@xen.org>

> > XENMEM_add_to_physmap_gmfn_range is like XENMEM_add_to_physmap on
> > the gmfn space but the number of pages on which xen will iterate.
> > 
> > Use the 16 bits padding between .domid and .space in struct 
> > xen_add_to_physmap
> > to keep compatibility with older versions.
> > 
> > Signed-off-by: Jean Guyader <jean.guyader@eu.citrix.com>
> > ---
> >  xen/arch/x86/mm.c           |   22 +++++++++++++++++++++-
> >  xen/include/public/memory.h |    4 ++++
> >  2 files changed, 25 insertions(+), 1 deletions(-)
> 
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Re: [PATCH 4/6] mm: New XENMEM, XENMEM_add_to_physmap_gmfn_range

[Tim Deegan]

At 13:32 +0000 on 08 Nov (1320759148), Tim Deegan wrote:
> At 02:20 -0700 on 08 Nov (1320718835), Jan Beulich wrote:
> > >>> On 07.11.11 at 19:25, Jean Guyader <jean.guyader@eu.citrix.com> wrote:
> > 
> > Sorry, noticed this only now, but neither title nor description of this
> > are in sync with the actual patch.
> 
> 
> Indeed; they should be updated.  But otherwise:
> Acked-by: Tim Deegan <tim@xen.org>

No, wait, that was the other patch, which I already acked!
ENOCOFFEE. :(

I have a few other comments on this patch...

> diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
> index f75011e..cca64b8 100644
> --- a/xen/arch/x86/mm.c
> +++ b/xen/arch/x86/mm.c
> @@ -4714,9 +4714,29 @@ long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg)
>              return -EPERM;
>          }
>  
> -        xenmem_add_to_physmap(d, xatp);
> +        if ( xatp.space != XENMAPSPACE_gmfn_range )
> +            xatp.size = 1;
> +
> +        for ( ; xatp.size > 0; xatp.size-- )
> +        {
> +            if ( hypercall_preempt_check() )
> +            {
> +                rc = -EAGAIN;
> +                break;
> +            }
> +            xenmem_add_to_physmap(d, xatp);
> +            xatp.idx++;
> +            xatp.gpfn++;
> +        }

Having moved XATP into its own function, this loop probably belongs in
that function. 

While I'm looking at it, updating two loop vars explicitly and one in
the header is a bit ugly - why not just use a while() and update all
three together?


>          rcu_unlock_domain(d);
> +        if ( rc == -EAGAIN )
> +        {
> +            if ( copy_to_guest(arg, &xatp, 1) )
> +                return -EFAULT;
> +            rc = hypercall_create_continuation(
> +                    __HYPERVISOR_memory_op, "ih", op, arg);
> +        }

I think this might need some compat glue in
arch/x86/x86_64/compat/mm.c for it to work with 32-bit callers.


Tim.