From: Anthony Liguori <anthony@codemonkey.ws>
To: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH] Introduce kvm logging interface.
Date: Mon, 10 Nov 2008 11:15:39 -0600 [thread overview]
Message-ID: <49186C3B.4070909@codemonkey.ws> (raw)
In-Reply-To: <1226342253-8887-4-git-send-email-glommer@redhat.com>
Glauber Costa wrote:
> Introduce functions to stop and start logging of memory regions.
> We select region based on its start address.
>
> Signed-off-by: Glauber Costa <glommer@redhat.com>
> ---
> kvm-all.c | 140 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> kvm.h | 5 ++
> 2 files changed, 145 insertions(+), 0 deletions(-)
>
> diff --git a/kvm-all.c b/kvm-all.c
> index 6d50609..c2c253f 100644
> --- a/kvm-all.c
> +++ b/kvm-all.c
> @@ -31,8 +31,15 @@
> do { } while (0)
> #endif
>
> +#define warning(fmt, ...) \
> + do { fprintf(stderr, "%s:%d" fmt, __func__, __LINE__, ## __VA_ARGS__); } while (0)
>
I'd rather this be a dprintf(). Error paths should not be chatty for
normal users because then a malicious can potentially start a crap-flood.
> typedef struct kvm_userspace_memory_region KVMSlot;
>
> +#define kvm_uaddr(addr) ((addr) + (ram_addr_t)phys_ram_base)
>
Should always be a static inline when possible, but I'm suspicious that
this function should even exist.
> +typedef struct kvm_dirty_log KVMDirty;
>
KVMDirtyLog?
> int kvm_allowed = 0;
>
> struct KVMState
> @@ -71,6 +78,24 @@ static KVMSlot *kvm_lookup_slot(KVMState *s, target_phys_addr_t start_addr)
> return NULL;
> }
>
> +/* find the slot correspondence using userspace_addr as a key */
> +static KVMSlot *kvm_lookup_slot_uaddr(KVMState *s, ram_addr_t addr)
> +{
> + int i;
> +
> + uint64_t uaddr = (uint64_t)kvm_uaddr(addr);
>
Since this is the only use of kvm_uaddr, you could just fold the #define
into here.
> +void kvm_physical_sync_dirty_bitmap(ram_addr_t start_addr, ram_addr_t end_addr)
> +{
> + KVMState *s = kvm_state;
> + KVMSlot *mem = kvm_lookup_slot_uaddr(s, start_addr);
> + KVMDirty d;
> + unsigned long alloc_size = mem->memory_size >> TARGET_PAGE_BITS / sizeof(d.dirty_bitmap);
> + ram_addr_t addr;
> +
> + dprintf("sync addr: %lx %llx %llx\n", start_addr, mem->guest_phys_addr, kvm_uaddr(start_addr));
> + if (mem == NULL) {
> + fprintf(stderr, "BUG: %s: invalid parameters\n", __func__);
> + return;
> + }
> +
> + d.dirty_bitmap = qemu_mallocz(alloc_size);
> +
> + if (d.dirty_bitmap == NULL) {
> + warning("Could not allocate dirty bitmap\n");
> + return;
> + }
> +
> + d.slot = mem->slot;
> + dprintf("slot %d, phys_addr %llx, uaddr: %llx\n",
> + d.slot, mem->guest_phys_addr, mem->userspace_addr);
> +
> + if (kvm_vm_ioctl(s, KVM_GET_DIRTY_LOG, &d) == -1) {
> + warning("ioctl failed %d\n", errno);
> + goto out;
> + }
> +
> + for (addr = start_addr; addr < end_addr; addr += TARGET_PAGE_SIZE){
> + unsigned nr;
> + nr = (uint32_t)((uint64_t)kvm_uaddr(addr) - mem->userspace_addr) >> TARGET_PAGE_BITS;
> + phys_ram_dirty[addr >> TARGET_PAGE_BITS] |= lookup_bitmap_phys((unsigned long *)d.dirty_bitmap, nr);
>
This is only setting the VGA_DIRTY_FLAG and really only by happy
coincidence. If that was on purpose, then shame on you ;-) You need to
set all bits in order for live migration to work.
You should use cpu_physical_memory_set_dirty(). I also think you could
simplify things by folding lookup_bitmap_phys() into this function but
that's not a requirement for merging.
> + }
> +out:
> + qemu_free(d.dirty_bitmap);
> +}
> +
> int kvm_init(int smp_cpus)
> {
> KVMState *s;
> diff --git a/kvm.h b/kvm.h
> index 37102b4..39e9048 100644
> --- a/kvm.h
> +++ b/kvm.h
> @@ -38,6 +38,11 @@ void kvm_set_phys_mem(target_phys_addr_t start_addr,
> ram_addr_t size,
> ram_addr_t phys_offset);
>
> +int kvm_physical_memory_get_dirty(ram_addr_t addr);
>
This function no longer exists.
If you fix these issues, then I think it's ready for merging.
Regards,
Anthony Liguori
> +void kvm_physical_sync_dirty_bitmap(ram_addr_t start_addr, ram_addr_t end_addr);
> +
> +int kvm_log_start(target_phys_addr_t phys_addr, target_phys_addr_t len);
> +int kvm_log_stop(target_phys_addr_t phys_addr, target_phys_addr_t len);
> /* internal API */
>
> struct KVMState;
>
next prev parent reply other threads:[~2008-11-10 17:15 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-11-10 18:37 [Qemu-devel] [PATCH 0/4] VGA optimization Glauber Costa
2008-11-10 16:44 ` Glauber Costa
2008-11-10 18:37 ` [Qemu-devel] [PATCH] move vga_io_address to VGA State Glauber Costa
2008-11-10 18:37 ` [Qemu-devel] [PATCH] de-register mem region for MMIO Glauber Costa
2008-11-10 18:37 ` [Qemu-devel] [PATCH] Introduce kvm logging interface Glauber Costa
2008-11-10 17:15 ` Anthony Liguori [this message]
2008-11-10 17:24 ` Anthony Liguori
2008-11-10 18:37 ` [Qemu-devel] [PATCH] vga optimization Glauber Costa
2008-11-10 18:39 ` [Qemu-devel] [PATCH 0/4] VGA optimization Anthony Liguori
2008-11-10 18:47 ` Anthony Liguori
2008-11-10 18:50 ` Glauber Costa
2008-11-10 20:17 ` Glauber Costa
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=49186C3B.4070909@codemonkey.ws \
--to=anthony@codemonkey.ws \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).