dfaggioli/qemu
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

migration: fix cgs-tdx to support migration cancel

Browse Source 
Set nr_channels at migration runtime
Pass the gfn_end to KVM, no need to align on blocked size.

Signed-off-by: Wei Wang <wei.w.wang@intel.com>
This commit is contained in:
Wei Wang
2023-02-19 20:13:13 +08:00
committed by Nikolay Borisov
parent 7e81998399
commit b581c14b8f
4 changed files with 28 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
migration/ram.c
View File

@@ -2538,8 +2538,7 @@ void ram_save_cancel(void)
RAMBlock *block = ram_state->last_seen_block;
unsigned long page = ram_state->last_page;
ram_addr_t offset = ((ram_addr_t)page) << TARGET_PAGE_BITS;
uint64_t align_size = 1UL << block->clear_bmap_shift;
hwaddr gpa, gfn, gfn_aligned;
hwaddr gpa, gfn;
int ret;
if (!ram_counters.cgs_epochs) {
@@ -2557,13 +2556,7 @@ void ram_save_cancel(void)
}
gfn = gpa >> TARGET_PAGE_BITS;
/*
* Pages that have been cleared in the clear log are write-protected, and
* need to be restored. So make it aligned on the clear_bmap boundary.
*/
gfn_aligned = QEMU_ALIGN_UP(gfn, align_size);
ret = cgs_mig_savevm_state_ram_cancel(ram_state->f, gfn_aligned);
ret = cgs_mig_savevm_state_ram_cancel(ram_state->f, gfn);
if (ret) {
error_report("%s failed: %s", __func__, strerror(ret));
}