system/physmem: make ram_block_discard_range() handle guest_memfd
Most callers of ram_block_discard_range() want to discard both the shared and guest_memfd backing. Only kvm_convert_memory() intentionally discards a single plane during private/shared conversions. Rename the current implementation to ram_block_discard_shared_range() and make ram_block_discard_range() a composite that also discards guest_memfd when present (rb->guest_memfd >= 0). This ensures callers like virtio-mem, virtio-balloon, hv-balloon, migration.. reclaim private pages on discard. Update kvm_convert_memory() to use the plane-specific ram_block_discard_shared_range() since it only needs to discard the shared backing when converting to private. Likewise, after TDVF image copy, use ram_block_discard_shared_range(). Reviewed-by: Peter Xu <peterx@redhat.com> Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Link: https://lore.kernel.org/r/20260604-rdm5-v5-11-5768e6a0943d@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
Marc-André Lureau committed
Jun 4, 2026 at 17:43 UTC
1f3241bcaeee0df47f72edf6abad71fe36f3b0e7
5 files changed
+26
-8
accel/kvm/kvm-all.c
+1
-1
@@ -3422,7 +3422,7 @@ int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
3422
*/
3423
goto out_unref;
3424
}
3425
- ret = ram_block_discard_range(rb, offset, size);
3425
+ ret = ram_block_discard_shared_range(rb, offset, size);
3426
} else {
3427
ret = ram_block_discard_guest_memfd_range(rb, offset, size);
3428
}
include/system/ramblock.h
+2
-1
@@ -103,7 +103,8 @@ struct RamBlockAttributes {
103
104
/* @offset: the offset within the RAMBlock */
105
int ram_block_discard_range(RAMBlock *rb, uint64_t offset, size_t length);
106
-/* @offset: the offset within the RAMBlock */
106
+int ram_block_discard_shared_range(RAMBlock *rb, uint64_t offset,
107
+ size_t length);
108
int ram_block_discard_guest_memfd_range(RAMBlock *rb, uint64_t offset,
109
size_t length);
110
system/physmem.c
+21
-4
@@ -4093,7 +4093,7 @@ int qemu_ram_foreach_block(RAMBlockIterFunc func, void *opaque)
4093
* Returns: 0 on success, none-0 on failure
4094
*
4095
*/
4096
-int ram_block_discard_range(RAMBlock *rb, uint64_t offset, size_t length)
4096
+int ram_block_discard_shared_range(RAMBlock *rb, uint64_t offset, size_t length)
4097
{
4098
int ret = -1;
4099
@@ -4142,7 +4142,7 @@ int ram_block_discard_range(RAMBlock *rb, uint64_t offset, size_t length)
4142
* have a MAP_PRIVATE mapping, possibly messing with other
4143
* MAP_PRIVATE/MAP_SHARED mappings. There is no easy way to
4144
* change that behavior whithout violating the promised
4145
- * semantics of ram_block_discard_range().
4145
+ * semantics of ram_block_discard_shared_range().
4146
*
4147
* Only warn, because it works as long as nobody else uses that
4148
* file.
@@ -4198,8 +4198,9 @@ int ram_block_discard_range(RAMBlock *rb, uint64_t offset, size_t length)
4198
goto err;
4199
#endif
4200
}
4201
- trace_ram_block_discard_range(rb->idstr, host_startaddr, length,
4202
- need_madvise, need_fallocate, ret);
4201
+ trace_ram_block_discard_shared_range(rb->idstr, host_startaddr, length,
4202
+ need_madvise, need_fallocate,
4203
+ ret);
4204
} else {
4205
error_report("%s: Overrun block '%s' (%" PRIu64 "/%zx/" RAM_ADDR_FMT")",
4206
__func__, rb->idstr, offset, length, rb->max_length);
@@ -4209,6 +4210,22 @@ err:
4210
return ret;
4211
}
4212
4213
+int ram_block_discard_range(RAMBlock *rb, uint64_t offset, size_t length)
4214
+{
4215
+ int ret;
4216
+
4217
+ ret = ram_block_discard_shared_range(rb, offset, length);
4218
+ if (ret) {
4219
+ return ret;
4220
+ }
4221
+
4222
+ if (rb->guest_memfd >= 0) {
4223
+ ret = ram_block_discard_guest_memfd_range(rb, offset, length);
4224
+ }
4225
+
4226
+ return ret;
4227
+}
4228
+
4229
int ram_block_discard_guest_memfd_range(RAMBlock *rb, uint64_t offset,
4230
size_t length)
4231
{
system/trace-events
+1
-1
@@ -32,7 +32,7 @@ global_dirty_changed(unsigned int bitmask) "bitmask 0x%"PRIx32
32
address_space_map(void *as, uint64_t addr, uint64_t len, bool is_write, uint32_t attrs) "as:%p addr 0x%"PRIx64":%"PRIx64" write:%d attrs:0x%x"
33
find_ram_offset(uint64_t size, uint64_t offset) "size: 0x%" PRIx64 " @ 0x%" PRIx64
34
find_ram_offset_loop(uint64_t size, uint64_t candidate, uint64_t offset, uint64_t next, uint64_t mingap) "trying size: 0x%" PRIx64 " @ 0x%" PRIx64 ", offset: 0x%" PRIx64" next: 0x%" PRIx64 " mingap: 0x%" PRIx64
35
-ram_block_discard_range(const char *rbname, void *hva, size_t length, bool need_madvise, bool need_fallocate, int ret) "%s@%p + 0x%zx: madvise: %d fallocate: %d ret: %d"
35
+ram_block_discard_shared_range(const char *rbname, void *hva, size_t length, bool need_madvise, bool need_fallocate, int ret) "%s@%p + 0x%zx: madvise: %d fallocate: %d ret: %d"
36
qemu_ram_alloc_shared(const char *name, size_t size, size_t max_size, int fd, void *host) "%s size %zu max_size %zu fd %d host %p"
37
38
subpage_register(void *subpage, uint32_t start, uint32_t end, int idx, int eidx, uint16_t section) "subpage %p start 0x%08x end 0x%08x idx 0x%08x eidx 0x%08x section %u"
target/i386/kvm/tdx.c
+1
-1
@@ -385,7 +385,7 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
385
* KVM_MEMORY_MAPPING. It becomes useless.
386
*/
387
ram_block = tdx_guest->tdvf_mr->ram_block;
388
- ram_block_discard_range(ram_block, 0, ram_block->max_length);
388
+ ram_block_discard_shared_range(ram_block, 0, ram_block->max_length);
389
390
tdx_vm_ioctl(KVM_TDX_FINALIZE_VM, 0, NULL, &error_fatal);
391
CONFIDENTIAL_GUEST_SUPPORT(tdx_guest)->ready = true;