Skip to contentSkip to conversation

s390x/kvm: Properly detect page size of initial memory

PATCH Applied David Hildenbrand 4 participants 8 messages Apr 17, 2019 – Apr 25, 2019

lore
Share & tools
lore
#0 lore raw

v3 cover s390x/kvm: Properly detect page size of initial memory

While the detection on s390x of the maximum ram pagesize works by pure luck, we should try to do better. Especially, once "-mem-path" is replaced by memory backends, the current approach will break.

Configure the maximum page size after configuring initital memory. Provide a function to detect the maximum ram pagesize

I think this can be deferred to 4.1 as I don't think anything is actually broken right now.

v2 -> v3:
- Fix compile error
- Tested that hugepages correctly result in an error with "modprobe kvm hpage=0", but not with "modprobe kvm hpage=1".

v1 -> v2:
- Add a better comment describing that only initial memory is considered
- Add "exec: Introduce qemu_getmaxrampagesize() and rename qemu_getrampagesize()"

David Hildenbrand (2):
s390x/kvm: Configure page size after memory has actually been
initialized
exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c | 12 +++++++++++
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 target/s390x/cpu.c         |  7 ++++++
 target/s390x/cpu.h         |  1 +
 target/s390x/kvm-stub.c    |  4 ++++
 target/s390x/kvm.c         | 35 ++++++++++++------------------
 target/s390x/kvm_s390x.h   |  1 +
 11 files changed, 85 insertions(+), 30 deletions(-)
Signature · 1 line
2.20.1
#1 lore raw

PATCH v3 1/2 s390x/kvm: Configure page size after memory has actually been initialized

Right now we configure the pagesize quite early, when initializing KVM. This is long before system memory is actually allocated via memory_region_allocate_system_memory(), and therefore memory backends marked as mapped.

Instead, let's configure the maximum page size after initializing memory in s390_memory_init(). cap_hpage_1m is still properly configured before creating any CPUs, and therefore before configuring the CPU model and eventually enabling CMMA.

This is not a fix but rather a preparation for the future, when initial memory might reside on memory backends (not the case for s390x right now) We will replace qemu_getrampagesize() soon by a function that will always return the maximum page size (not the minimum page size, which only works by pure luck so far, as there are no memory backends).

---
 hw/s390x/s390-virtio-ccw.c | 12 ++++++++++++
 target/s390x/cpu.c         |  7 +++++++
 target/s390x/cpu.h         |  1 +
 target/s390x/kvm-stub.c    |  4 ++++
 target/s390x/kvm.c         | 35 ++++++++++++++---------------------
 target/s390x/kvm_s390x.h   |  1 +
 6 files changed, 39 insertions(+), 21 deletions(-)
hw/s390x/s390-virtio-ccw.c +12 -0
index d11069b860..3be5679657 100644
@@ -15,6 +15,7 @@
 #include "cpu.h"
 #include "hw/boards.h"
 #include "exec/address-spaces.h"
+#include "exec/ram_addr.h"
 #include "hw/s390x/s390-virtio-hcall.h"
 #include "hw/s390x/sclp.h"
 #include "hw/s390x/s390_flic.h"
@@ -163,6 +164,7 @@ static void s390_memory_init(ram_addr_t mem_size)
     MemoryRegion *sysmem = get_system_memory();
     ram_addr_t chunk, offset = 0;
     unsigned int number = 0;
+    Error *local_err = NULL;
     gchar *name;
 
     /* allocate RAM for core */
@@ -182,6 +184,15 @@ static void s390_memory_init(ram_addr_t mem_size)
     }
     g_free(name);
 
+    /*
+     * Configure the maximum page size. As no memory devices were created
+     * yet, this is the page size of initial memory only.
+     */
+    s390_set_max_pagesize(qemu_getrampagesize(), &local_err);
+    if (local_err) {
+        error_report_err(local_err);
+        exit(EXIT_FAILURE);
+    }
     /* Initialize storage key device */
     s390_skeys_init();
     /* Initialize storage attributes device */
@@ -253,6 +264,7 @@ static void ccw_init(MachineState *machine)
     DeviceState *dev;
 
     s390_sclp_init();
+    /* init memory + setup max page size. Required for the CPU model */
     s390_memory_init(machine->ram_size);
 
     /* init CPUs (incl. CPU model) early so s390_has_feature() works */
target/s390x/cpu.c +7 -0
index 698dd9cb82..b58ef0a8ef 100644
@@ -399,6 +399,13 @@ int s390_set_memory_limit(uint64_t new_limit, uint64_t *hw_limit)
     return 0;
 }
 
+void s390_set_max_pagesize(uint64_t pagesize, Error **errp)
+{
+    if (kvm_enabled()) {
+        kvm_s390_set_max_pagesize(pagesize, errp);
+    }
+}
+
 void s390_cmma_reset(void)
 {
     if (kvm_enabled()) {
target/s390x/cpu.h +1 -0
index cb6d77053a..c14be2b5ba 100644
@@ -734,6 +734,7 @@ static inline void s390_do_cpu_load_normal(CPUState *cs, run_on_cpu_data arg)
 /* cpu.c */
 void s390_crypto_reset(void);
 int s390_set_memory_limit(uint64_t new_limit, uint64_t *hw_limit);
+void s390_set_max_pagesize(uint64_t pagesize, Error **errp);
 void s390_cmma_reset(void);
 void s390_enable_css_support(S390CPU *cpu);
 int s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch_id,
target/s390x/kvm-stub.c +4 -0
index bf7795e47a..22b4514ca6 100644
@@ -93,6 +93,10 @@ int kvm_s390_set_mem_limit(uint64_t new_limit, uint64_t *hw_limit)
     return 0;
 }
 
+void kvm_s390_set_max_pagesize(uint64_t pagesize, Error **errp)
+{
+}
+
 void kvm_s390_crypto_reset(void)
 {
 }
target/s390x/kvm.c +14 -21
index 19530fb94e..bee73dc1a4 100644
@@ -283,44 +283,37 @@ void kvm_s390_crypto_reset(void)
     }
 }
 
-static int kvm_s390_configure_mempath_backing(KVMState *s)
+void kvm_s390_set_max_pagesize(uint64_t pagesize, Error **errp)
 {
-    size_t path_psize = qemu_getrampagesize();
-
-    if (path_psize == 4 * KiB) {
-        return 0;
+    if (pagesize == 4 * KiB) {
+        return;
     }
 
     if (!hpage_1m_allowed()) {
-        error_report("This QEMU machine does not support huge page "
-                     "mappings");
-        return -EINVAL;
+        error_setg(errp, "This QEMU machine does not support huge page "
+                   "mappings");
+        return;
     }
 
-    if (path_psize != 1 * MiB) {
-        error_report("Memory backing with 2G pages was specified, "
-                     "but KVM does not support this memory backing");
-        return -EINVAL;
+    if (pagesize != 1 * MiB) {
+        error_setg(errp, "Memory backing with 2G pages was specified, "
+                   "but KVM does not support this memory backing");
+        return;
     }
 
-    if (kvm_vm_enable_cap(s, KVM_CAP_S390_HPAGE_1M, 0)) {
-        error_report("Memory backing with 1M pages was specified, "
-                     "but KVM does not support this memory backing");
-        return -EINVAL;
+    if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_HPAGE_1M, 0)) {
+        error_setg(errp, "Memory backing with 1M pages was specified, "
+                   "but KVM does not support this memory backing");
+        return;
     }
 
     cap_hpage_1m = 1;
-    return 0;
 }
 
 int kvm_arch_init(MachineState *ms, KVMState *s)
 {
     MachineClass *mc = MACHINE_GET_CLASS(ms);
 
-    if (kvm_s390_configure_mempath_backing(s)) {
-        return -EINVAL;
-    }
-
     mc->default_cpu_type = S390_CPU_TYPE_NAME("host");
     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
target/s390x/kvm_s390x.h +1 -0
index 6e52287da3..caf985955b 100644
@@ -36,6 +36,7 @@ int kvm_s390_cmma_active(void);
 void kvm_s390_cmma_reset(void);
 void kvm_s390_reset_vcpu(S390CPU *cpu);
 int kvm_s390_set_mem_limit(uint64_t new_limit, uint64_t *hw_limit);
+void kvm_s390_set_max_pagesize(uint64_t pagesize, Error **errp);
 void kvm_s390_crypto_reset(void);
 void kvm_s390_restart_interrupt(S390CPU *cpu);
 void kvm_s390_stop_interrupt(S390CPU *cpu);
Signature · 1 line
2.20.1
#2 lore raw

PATCH v3 2/2 exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

In reply to David Hildenbrand

Rename qemu_getrampagesize() to qemu_minrampagesize(). While at it, properly rename find_max_supported_pagesize() to find_min_backend_pagesize().

s390x is actually interested into the maximum ram pagesize, so introduce and use qemu_maxrampagesize().

Add a TODO, indicating that looking at any mapped memory backends is not 100% correct in some cases.

---
 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c |  2 +-
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 6 files changed, 47 insertions(+), 10 deletions(-)
exec.c +40 -4
index 6ab62f4eee..f9b834ef82 100644
@@ -1687,7 +1687,7 @@ void ram_block_dump(Monitor *mon)
  * when we actually open and map them.  Iterate over the file
  * descriptors instead, and use qemu_fd_getpagesize().
  */
-static int find_max_supported_pagesize(Object *obj, void *opaque)
+static int find_min_backend_pagesize(Object *obj, void *opaque)
 {
     long *hpsize_min = opaque;
 
@@ -1703,7 +1703,27 @@ static int find_max_supported_pagesize(Object *obj, void *opaque)
     return 0;
 }
 
-long qemu_getrampagesize(void)
+static int find_max_backend_pagesize(Object *obj, void *opaque)
+{
+    long *hpsize_max = opaque;
+
+    if (object_dynamic_cast(obj, TYPE_MEMORY_BACKEND)) {
+        HostMemoryBackend *backend = MEMORY_BACKEND(obj);
+        long hpsize = host_memory_backend_pagesize(backend);
+
+        if (host_memory_backend_is_mapped(backend) && (hpsize > *hpsize_max)) {
+            *hpsize_max = hpsize;
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * TODO: We assume right now that all mapped host memory backends are
+ * used as RAM, however some might be used for different purposes.
+ */
+long qemu_minrampagesize(void)
 {
     long hpsize = LONG_MAX;
     long mainrampagesize;
@@ -1723,7 +1743,7 @@ long qemu_getrampagesize(void)
      */
     memdev_root = object_resolve_path("/objects", NULL);
     if (memdev_root) {
-        object_child_foreach(memdev_root, find_max_supported_pagesize, &hpsize);
+        object_child_foreach(memdev_root, find_min_backend_pagesize, &hpsize);
     }
     if (hpsize == LONG_MAX) {
         /* No additional memory regions found ==> Report main RAM page size */
@@ -1746,8 +1766,24 @@ long qemu_getrampagesize(void)
 
     return hpsize;
 }
+
+long qemu_maxrampagesize(void)
+{
+    long pagesize = qemu_mempath_getpagesize(mem_path);
+    Object *memdev_root = object_resolve_path("/objects", NULL);
+
+    if (memdev_root) {
+        object_child_foreach(memdev_root, find_max_backend_pagesize,
+                             &pagesize);
+    }
+    return pagesize;
+}
 #else
-long qemu_getrampagesize(void)
+long qemu_minrampagesize(void)
+{
+    return getpagesize();
+}
+long qemu_maxrampagesize(void)
 {
     return getpagesize();
 }
hw/ppc/spapr_caps.c +2 -2
index edc5ed0e0c..9b1c10baa6 100644
@@ -347,7 +347,7 @@ static void cap_hpt_maxpagesize_apply(SpaprMachineState *spapr,
         warn_report("Many guests require at least 64kiB hpt-max-page-size");
     }
 
-    spapr_check_pagesize(spapr, qemu_getrampagesize(), errp);
+    spapr_check_pagesize(spapr, qemu_minrampagesize(), errp);
 }
 
 static bool spapr_pagesize_cb(void *opaque, uint32_t seg_pshift,
@@ -609,7 +609,7 @@ static SpaprCapabilities default_caps_with_cpu(SpaprMachineState *spapr,
         uint8_t mps;
 
         if (kvmppc_hpt_needs_host_contiguous_pages()) {
-            mps = ctz64(qemu_getrampagesize());
+            mps = ctz64(qemu_minrampagesize());
         } else {
             mps = 34; /* allow everything up to 16GiB, i.e. everything */
         }
hw/s390x/s390-virtio-ccw.c +1 -1
index 3be5679657..7e256d3d31 100644
@@ -188,7 +188,7 @@ static void s390_memory_init(ram_addr_t mem_size)
      * Configure the maximum page size. As no memory devices were created
      * yet, this is the page size of initial memory only.
      */
-    s390_set_max_pagesize(qemu_getrampagesize(), &local_err);
+    s390_set_max_pagesize(qemu_maxrampagesize(), &local_err);
     if (local_err) {
         error_report_err(local_err);
         exit(EXIT_FAILURE);
hw/vfio/spapr.c +1 -1
index 57fe758e54..96c0ad9d9b 100644
@@ -148,7 +148,7 @@ int vfio_spapr_create_window(VFIOContainer *container,
     uint64_t pagesize = memory_region_iommu_get_min_page_size(iommu_mr);
     unsigned entries, bits_total, bits_per_level, max_levels;
     struct vfio_iommu_spapr_tce_create create = { .argsz = sizeof(create) };
-    long rampagesize = qemu_getrampagesize();
+    long rampagesize = qemu_minrampagesize();
 
     /*
      * The host might not support the guest supported IOMMU page size,
include/exec/ram_addr.h +2 -1
index 9ecd911c3e..139ad79390 100644
@@ -73,7 +73,8 @@ static inline unsigned long int ramblock_recv_bitmap_offset(void *host_addr,
 
 bool ramblock_is_pmem(RAMBlock *rb);
 
-long qemu_getrampagesize(void);
+long qemu_minrampagesize(void);
+long qemu_maxrampagesize(void);
 
 /**
  * qemu_ram_alloc_from_file,
target/ppc/kvm.c +1 -1
index 2427c8ee13..59d92c4275 100644
@@ -2136,7 +2136,7 @@ uint64_t kvmppc_rma_size(uint64_t current_size, unsigned int hash_shift)
     /* Find the largest hardware supported page size that's less than
      * or equal to the (logical) backing page size of guest RAM */
     kvm_get_smmu_info(&info, &error_fatal);
-    rampagesize = qemu_getrampagesize();
+    rampagesize = qemu_minrampagesize();
     best_page_shift = 0;
 
     for (i = 0; i < KVM_PPC_PAGE_SIZES_MAX_SZ; i++) {
Signature · 1 line
2.20.1
Reviewed-by on v3 2/2
#3 lore raw

Re: [Qemu-devel] [PATCH v3 2/2] exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

Quoted 11 lines · David Hildenbrand

Rename qemu_getrampagesize() to qemu_minrampagesize(). While at it, properly rename find_max_supported_pagesize() to find_min_backend_pagesize().

s390x is actually interested into the maximum ram pagesize, so introduce and use qemu_maxrampagesize().

Add a TODO, indicating that looking at any mapped memory backends is not 100% correct in some cases.

Quoted 161 lines
---
 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c |  2 +-
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 6 files changed, 47 insertions(+), 10 deletions(-)
exec.c +40 -4
index 6ab62f4eee..f9b834ef82 100644
@@ -1687,7 +1687,7 @@ void ram_block_dump(Monitor *mon)
  * when we actually open and map them.  Iterate over the file
  * descriptors instead, and use qemu_fd_getpagesize().
  */
-static int find_max_supported_pagesize(Object *obj, void *opaque)
+static int find_min_backend_pagesize(Object *obj, void *opaque)
 {
     long *hpsize_min = opaque;
 
@@ -1703,7 +1703,27 @@ static int find_max_supported_pagesize(Object *obj, void *opaque)
     return 0;
 }
 
-long qemu_getrampagesize(void)
+static int find_max_backend_pagesize(Object *obj, void *opaque)
+{
+    long *hpsize_max = opaque;
+
+    if (object_dynamic_cast(obj, TYPE_MEMORY_BACKEND)) {
+        HostMemoryBackend *backend = MEMORY_BACKEND(obj);
+        long hpsize = host_memory_backend_pagesize(backend);
+
+        if (host_memory_backend_is_mapped(backend) && (hpsize > *hpsize_max)) {
+            *hpsize_max = hpsize;
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * TODO: We assume right now that all mapped host memory backends are
+ * used as RAM, however some might be used for different purposes.
+ */
+long qemu_minrampagesize(void)
 {
     long hpsize = LONG_MAX;
     long mainrampagesize;
@@ -1723,7 +1743,7 @@ long qemu_getrampagesize(void)
      */
     memdev_root = object_resolve_path("/objects", NULL);
     if (memdev_root) {
-        object_child_foreach(memdev_root, find_max_supported_pagesize, &hpsize);
+        object_child_foreach(memdev_root, find_min_backend_pagesize, &hpsize);
     }
     if (hpsize == LONG_MAX) {
         /* No additional memory regions found ==> Report main RAM page size */
@@ -1746,8 +1766,24 @@ long qemu_getrampagesize(void)
 
     return hpsize;
 }
+
+long qemu_maxrampagesize(void)
+{
+    long pagesize = qemu_mempath_getpagesize(mem_path);
+    Object *memdev_root = object_resolve_path("/objects", NULL);
+
+    if (memdev_root) {
+        object_child_foreach(memdev_root, find_max_backend_pagesize,
+                             &pagesize);
+    }
+    return pagesize;
+}
 #else
-long qemu_getrampagesize(void)
+long qemu_minrampagesize(void)
+{
+    return getpagesize();
+}
+long qemu_maxrampagesize(void)
 {
     return getpagesize();
 }
hw/ppc/spapr_caps.c +2 -2
index edc5ed0e0c..9b1c10baa6 100644
@@ -347,7 +347,7 @@ static void cap_hpt_maxpagesize_apply(SpaprMachineState *spapr,
         warn_report("Many guests require at least 64kiB hpt-max-page-size");
     }
 
-    spapr_check_pagesize(spapr, qemu_getrampagesize(), errp);
+    spapr_check_pagesize(spapr, qemu_minrampagesize(), errp);
 }
 
 static bool spapr_pagesize_cb(void *opaque, uint32_t seg_pshift,
@@ -609,7 +609,7 @@ static SpaprCapabilities default_caps_with_cpu(SpaprMachineState *spapr,
         uint8_t mps;
 
         if (kvmppc_hpt_needs_host_contiguous_pages()) {
-            mps = ctz64(qemu_getrampagesize());
+            mps = ctz64(qemu_minrampagesize());
         } else {
             mps = 34; /* allow everything up to 16GiB, i.e. everything */
         }
hw/s390x/s390-virtio-ccw.c +1 -1
index 3be5679657..7e256d3d31 100644
@@ -188,7 +188,7 @@ static void s390_memory_init(ram_addr_t mem_size)
      * Configure the maximum page size. As no memory devices were created
      * yet, this is the page size of initial memory only.
      */
-    s390_set_max_pagesize(qemu_getrampagesize(), &local_err);
+    s390_set_max_pagesize(qemu_maxrampagesize(), &local_err);
     if (local_err) {
         error_report_err(local_err);
         exit(EXIT_FAILURE);
hw/vfio/spapr.c +1 -1
index 57fe758e54..96c0ad9d9b 100644
@@ -148,7 +148,7 @@ int vfio_spapr_create_window(VFIOContainer *container,
     uint64_t pagesize = memory_region_iommu_get_min_page_size(iommu_mr);
     unsigned entries, bits_total, bits_per_level, max_levels;
     struct vfio_iommu_spapr_tce_create create = { .argsz = sizeof(create) };
-    long rampagesize = qemu_getrampagesize();
+    long rampagesize = qemu_minrampagesize();
 
     /*
      * The host might not support the guest supported IOMMU page size,
include/exec/ram_addr.h +2 -1
index 9ecd911c3e..139ad79390 100644
@@ -73,7 +73,8 @@ static inline unsigned long int ramblock_recv_bitmap_offset(void *host_addr,
 
 bool ramblock_is_pmem(RAMBlock *rb);
 
-long qemu_getrampagesize(void);
+long qemu_minrampagesize(void);
+long qemu_maxrampagesize(void);
 
 /**
  * qemu_ram_alloc_from_file,
target/ppc/kvm.c +1 -1
index 2427c8ee13..59d92c4275 100644
@@ -2136,7 +2136,7 @@ uint64_t kvmppc_rma_size(uint64_t current_size, unsigned int hash_shift)
     /* Find the largest hardware supported page size that's less than
      * or equal to the (logical) backing page size of guest RAM */
     kvm_get_smmu_info(&info, &error_fatal);
-    rampagesize = qemu_getrampagesize();
+    rampagesize = qemu_minrampagesize();
     best_page_shift = 0;
 
     for (i = 0; i < KVM_PPC_PAGE_SIZES_MAX_SZ; i++) {
Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
Reviewed-by on v3 2/2
#4 lore raw

Re: [Qemu-devel] [PATCH v3 2/2] exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

In reply to David Hildenbrand
Quoted 11 lines · On Wed, 17 Apr 2019 13:31:43 +0200 David Hildenbrand <[email protected]> wrote:

Rename qemu_getrampagesize() to qemu_minrampagesize(). While at it, properly rename find_max_supported_pagesize() to find_min_backend_pagesize().

s390x is actually interested into the maximum ram pagesize, so introduce and use qemu_maxrampagesize().

Add a TODO, indicating that looking at any mapped memory backends is not 100% correct in some cases.

Quoted 161 lines
---
 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c |  2 +-
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 6 files changed, 47 insertions(+), 10 deletions(-)
exec.c +40 -4
index 6ab62f4eee..f9b834ef82 100644
@@ -1687,7 +1687,7 @@ void ram_block_dump(Monitor *mon)
  * when we actually open and map them.  Iterate over the file
  * descriptors instead, and use qemu_fd_getpagesize().
  */
-static int find_max_supported_pagesize(Object *obj, void *opaque)
+static int find_min_backend_pagesize(Object *obj, void *opaque)
 {
     long *hpsize_min = opaque;
 
@@ -1703,7 +1703,27 @@ static int find_max_supported_pagesize(Object *obj, void *opaque)
     return 0;
 }
 
-long qemu_getrampagesize(void)
+static int find_max_backend_pagesize(Object *obj, void *opaque)
+{
+    long *hpsize_max = opaque;
+
+    if (object_dynamic_cast(obj, TYPE_MEMORY_BACKEND)) {
+        HostMemoryBackend *backend = MEMORY_BACKEND(obj);
+        long hpsize = host_memory_backend_pagesize(backend);
+
+        if (host_memory_backend_is_mapped(backend) && (hpsize > *hpsize_max)) {
+            *hpsize_max = hpsize;
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * TODO: We assume right now that all mapped host memory backends are
+ * used as RAM, however some might be used for different purposes.
+ */
+long qemu_minrampagesize(void)
 {
     long hpsize = LONG_MAX;
     long mainrampagesize;
@@ -1723,7 +1743,7 @@ long qemu_getrampagesize(void)
      */
     memdev_root = object_resolve_path("/objects", NULL);
     if (memdev_root) {
-        object_child_foreach(memdev_root, find_max_supported_pagesize, &hpsize);
+        object_child_foreach(memdev_root, find_min_backend_pagesize, &hpsize);
     }
     if (hpsize == LONG_MAX) {
         /* No additional memory regions found ==> Report main RAM page size */
@@ -1746,8 +1766,24 @@ long qemu_getrampagesize(void)
 
     return hpsize;
 }
+
+long qemu_maxrampagesize(void)
+{
+    long pagesize = qemu_mempath_getpagesize(mem_path);
+    Object *memdev_root = object_resolve_path("/objects", NULL);
+
+    if (memdev_root) {
+        object_child_foreach(memdev_root, find_max_backend_pagesize,
+                             &pagesize);
+    }
+    return pagesize;
+}
 #else
-long qemu_getrampagesize(void)
+long qemu_minrampagesize(void)
+{
+    return getpagesize();
+}
+long qemu_maxrampagesize(void)
 {
     return getpagesize();
 }
hw/ppc/spapr_caps.c +2 -2
index edc5ed0e0c..9b1c10baa6 100644
@@ -347,7 +347,7 @@ static void cap_hpt_maxpagesize_apply(SpaprMachineState *spapr,
         warn_report("Many guests require at least 64kiB hpt-max-page-size");
     }
 
-    spapr_check_pagesize(spapr, qemu_getrampagesize(), errp);
+    spapr_check_pagesize(spapr, qemu_minrampagesize(), errp);
 }
 
 static bool spapr_pagesize_cb(void *opaque, uint32_t seg_pshift,
@@ -609,7 +609,7 @@ static SpaprCapabilities default_caps_with_cpu(SpaprMachineState *spapr,
         uint8_t mps;
 
         if (kvmppc_hpt_needs_host_contiguous_pages()) {
-            mps = ctz64(qemu_getrampagesize());
+            mps = ctz64(qemu_minrampagesize());
         } else {
             mps = 34; /* allow everything up to 16GiB, i.e. everything */
         }
hw/s390x/s390-virtio-ccw.c +1 -1
index 3be5679657..7e256d3d31 100644
@@ -188,7 +188,7 @@ static void s390_memory_init(ram_addr_t mem_size)
      * Configure the maximum page size. As no memory devices were created
      * yet, this is the page size of initial memory only.
      */
-    s390_set_max_pagesize(qemu_getrampagesize(), &local_err);
+    s390_set_max_pagesize(qemu_maxrampagesize(), &local_err);
     if (local_err) {
         error_report_err(local_err);
         exit(EXIT_FAILURE);
hw/vfio/spapr.c +1 -1
index 57fe758e54..96c0ad9d9b 100644
@@ -148,7 +148,7 @@ int vfio_spapr_create_window(VFIOContainer *container,
     uint64_t pagesize = memory_region_iommu_get_min_page_size(iommu_mr);
     unsigned entries, bits_total, bits_per_level, max_levels;
     struct vfio_iommu_spapr_tce_create create = { .argsz = sizeof(create) };
-    long rampagesize = qemu_getrampagesize();
+    long rampagesize = qemu_minrampagesize();
 
     /*
      * The host might not support the guest supported IOMMU page size,
include/exec/ram_addr.h +2 -1
index 9ecd911c3e..139ad79390 100644
@@ -73,7 +73,8 @@ static inline unsigned long int ramblock_recv_bitmap_offset(void *host_addr,
 
 bool ramblock_is_pmem(RAMBlock *rb);
 
-long qemu_getrampagesize(void);
+long qemu_minrampagesize(void);
+long qemu_maxrampagesize(void);
 
 /**
  * qemu_ram_alloc_from_file,
target/ppc/kvm.c +1 -1
index 2427c8ee13..59d92c4275 100644
@@ -2136,7 +2136,7 @@ uint64_t kvmppc_rma_size(uint64_t current_size, unsigned int hash_shift)
     /* Find the largest hardware supported page size that's less than
      * or equal to the (logical) backing page size of guest RAM */
     kvm_get_smmu_info(&info, &error_fatal);
-    rampagesize = qemu_getrampagesize();
+    rampagesize = qemu_minrampagesize();
     best_page_shift = 0;
 
     for (i = 0; i < KVM_PPC_PAGE_SIZES_MAX_SZ; i++) {
#5 lore raw

Re: [Qemu-devel] [PATCH v3 2/2] exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

In reply to David Hildenbrand
Quoted 11 lines · On Wed, 17 Apr 2019 13:31:43 +0200 David Hildenbrand <[email protected]> wrote:

Rename qemu_getrampagesize() to qemu_minrampagesize(). While at it, properly rename find_max_supported_pagesize() to find_min_backend_pagesize().

s390x is actually interested into the maximum ram pagesize, so introduce and use qemu_maxrampagesize().

Add a TODO, indicating that looking at any mapped memory backends is not 100% correct in some cases.

Quoted 8 lines
---
 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c |  2 +-
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 6 files changed, 47 insertions(+), 10 deletions(-)

This now has two R-bs, so unless somebody speaks up against it, I'll just take this through the s390x tree together with patch 1.

#6 lore raw
In reply to David Hildenbrand
Quoted 9 lines · On Wed, 17 Apr 2019 13:31:41 +0200 David Hildenbrand <[email protected]> wrote:

While the detection on s390x of the maximum ram pagesize works by pure luck, we should try to do better. Especially, once "-mem-path" is replaced by memory backends, the current approach will break.

Configure the maximum page size after configuring initital memory. Provide a function to detect the maximum ram pagesize

I think this can be deferred to 4.1 as I don't think anything is actually broken right now.

I think it is certainly 4.1 at that point in time ;)

Who is going to queue this? If not me, I'll try to find some time to review this next week.

Quoted 29 lines

v2 -> v3:
- Fix compile error
- Tested that hugepages correctly result in an error with "modprobe kvm hpage=0", but not with "modprobe kvm hpage=1".

v1 -> v2:
- Add a better comment describing that only initial memory is considered
- Add "exec: Introduce qemu_getmaxrampagesize() and rename qemu_getrampagesize()"

David Hildenbrand (2):
s390x/kvm: Configure page size after memory has actually been
initialized
exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c | 12 +++++++++++
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 target/s390x/cpu.c         |  7 ++++++
 target/s390x/cpu.h         |  1 +
 target/s390x/kvm-stub.c    |  4 ++++
 target/s390x/kvm.c         | 35 ++++++++++++------------------
 target/s390x/kvm_s390x.h   |  1 +
 11 files changed, 85 insertions(+), 30 deletions(-)
#7 lore raw
In reply to David Hildenbrand
Quoted 25 lines · On Wed, 17 Apr 2019 13:31:41 +0200 David Hildenbrand <[email protected]> wrote:

While the detection on s390x of the maximum ram pagesize works by pure luck, we should try to do better. Especially, once "-mem-path" is replaced by memory backends, the current approach will break.

Configure the maximum page size after configuring initital memory. Provide a function to detect the maximum ram pagesize

I think this can be deferred to 4.1 as I don't think anything is actually broken right now.

v2 -> v3:
- Fix compile error
- Tested that hugepages correctly result in an error with "modprobe kvm hpage=0", but not with "modprobe kvm hpage=1".

v1 -> v2:
- Add a better comment describing that only initial memory is considered
- Add "exec: Introduce qemu_getmaxrampagesize() and rename qemu_getrampagesize()"

David Hildenbrand (2):
s390x/kvm: Configure page size after memory has actually been
initialized
exec: Introduce qemu_maxrampagesize() and rename qemu_getrampagesize()

Quoted 13 lines
 exec.c                     | 44 ++++++++++++++++++++++++++++++++++----
 hw/ppc/spapr_caps.c        |  4 ++--
 hw/s390x/s390-virtio-ccw.c | 12 +++++++++++
 hw/vfio/spapr.c            |  2 +-
 include/exec/ram_addr.h    |  3 ++-
 target/ppc/kvm.c           |  2 +-
 target/s390x/cpu.c         |  7 ++++++
 target/s390x/cpu.h         |  1 +
 target/s390x/kvm-stub.c    |  4 ++++
 target/s390x/kvm.c         | 35 ++++++++++++------------------
 target/s390x/kvm_s390x.h   |  1 +
 11 files changed, 85 insertions(+), 30 deletions(-)

Thanks, applied.