From: Igor Mammedov <imammedo@redhat.com>
To: Vasilis Liaskovitis <vasilis.liaskovitis@profitbricks.com>
Cc: thilo.fromm@profitbricks.com, seabios@seabios.org,
qemu-devel@nongnu.org, mst@redhat.com
Subject: Re: [Qemu-devel] [RFC PATCH v2] i386: Add _PXM ACPI method to CPU objects
Date: Wed, 27 Nov 2013 16:53:30 +0100 [thread overview]
Message-ID: <20131127165330.075dcf41@nial.usersys.redhat.com> (raw)
In-Reply-To: <1385557363-25756-1-git-send-email-vasilis.liaskovitis@profitbricks.com>
On Wed, 27 Nov 2013 14:02:43 +0100
Vasilis Liaskovitis <vasilis.liaskovitis@profitbricks.com> wrote:
> This patch adds a _PXM method to ACPI CPU objects for the pc machine. The _PXM
> value is derived from the passed in guest info, same way as CPU SRAT entries.
>
> Currently, CPU SRAT entries are only enabled for cpus that are already present
> in the system. The SRAT entries for hotpluggable processors are disabled (flags
> bit 0 set to 0 in hw/i385/acpi-build.c:build_srat). Section 5.2.16.1 of ACPI
> spec mentions "If the Local APIC ID of a dynamically added processor is not
> present in the SRAT, a _PXM object must exist for the processor’s device or one
> of its ancestors in the ACPI Namespace." Since SRAT entries are not available
> for the hot-pluggable processors, a _PXM method must exist for them. Otherwise,
> the CPU is hot-added in the wrong NUMA node (default node 0).
>
> Even if CPU SRAT entries are enabled, _PXM method is what the linux kernel
> consults on hot-add time. Section 17.2.1 of ACPI spec mentions " OSPM will
> consume the SRAT only at boot time. OSPM should use _PXM for any devices that
> are hot-added into the system after boot up." To be more precise if SRAT
> information is available to the guest kernel, it is used. However, parsed SRAT
> info is reset and lost after hot-remove operations, see kernel commit c4c60524.
> This means that on a hot-unplug / hot-replug scenario, and without a _PXM
> method, the kernel may put a CPU on different nodes because SRAT info has been
> reset by a previous hot-remove operation.
>
> The above hot-remove/hot-add scenario has been tested on master, plus cpu-del
> patches from:
> https://lists.gnu.org/archive/html/qemu-devel/2013-10/msg01085.html
> With the curret _PXM patch, hot-added CPUs are always placed into the correct
> NUMA node, regardless of kernel behaviour.
>
> v1->v2:
> Make method return a DWORD integer
> Tested on qemu master + cpu-del patches
>
> Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis@profitbricks.com>
> Reviewed-by: Thilo Fromm <t-lo@thilo-fromm.de>
Patch looks good,
Please add patch to update hw/i386/ssdt-proc.hex.generated for hosts without iasl
for completness
>
> ---
> hw/i386/acpi-build.c | 5 +++++
> hw/i386/ssdt-proc.dsl | 5 +++++
> 2 files changed, 10 insertions(+)
>
> diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> index d089e1e..3c11ddc 100644
> --- a/hw/i386/acpi-build.c
> +++ b/hw/i386/acpi-build.c
> @@ -603,6 +603,7 @@ static inline char acpi_get_hex(uint32_t val)
> #define ACPI_PROC_OFFSET_CPUHEX (*ssdt_proc_name - *ssdt_proc_start + 2)
> #define ACPI_PROC_OFFSET_CPUID1 (*ssdt_proc_name - *ssdt_proc_start + 4)
> #define ACPI_PROC_OFFSET_CPUID2 (*ssdt_proc_id - *ssdt_proc_start)
> +#define ACPI_PROC_OFFSET_CPUPXM (*ssdt_proc_pxm - *ssdt_proc_start)
> #define ACPI_PROC_SIZEOF (*ssdt_proc_end - *ssdt_proc_start)
> #define ACPI_PROC_AML (ssdp_proc_aml + *ssdt_proc_start)
>
> @@ -724,6 +725,10 @@ build_ssdt(GArray *table_data, GArray *linker,
> proc[ACPI_PROC_OFFSET_CPUHEX+1] = acpi_get_hex(i);
> proc[ACPI_PROC_OFFSET_CPUID1] = i;
> proc[ACPI_PROC_OFFSET_CPUID2] = i;
> + proc[ACPI_PROC_OFFSET_CPUPXM] = guest_info->node_cpu[i];
> + proc[ACPI_PROC_OFFSET_CPUPXM + 1] = 0;
> + proc[ACPI_PROC_OFFSET_CPUPXM + 2] = 0;
> + proc[ACPI_PROC_OFFSET_CPUPXM + 3] = 0;
> }
>
> /* build this code:
> diff --git a/hw/i386/ssdt-proc.dsl b/hw/i386/ssdt-proc.dsl
> index 8229bfd..8d4c5bf 100644
> --- a/hw/i386/ssdt-proc.dsl
> +++ b/hw/i386/ssdt-proc.dsl
> @@ -47,6 +47,8 @@ DefinitionBlock ("ssdt-proc.aml", "SSDT", 0x01, "BXPC", "BXSSDT", 0x1)
> * also updating the C code.
> */
> Name(_HID, "ACPI0007")
> + ACPI_EXTRACT_NAME_DWORD_CONST ssdt_proc_pxm
> + Name(PXM, 0xAAAAAAAA)
> External(CPMA, MethodObj)
> External(CPST, MethodObj)
> External(CPEJ, MethodObj)
> @@ -59,5 +61,8 @@ DefinitionBlock ("ssdt-proc.aml", "SSDT", 0x01, "BXPC", "BXSSDT", 0x1)
> Method(_EJ0, 1, NotSerialized) {
> CPEJ(ID, Arg0)
> }
> + Method(_PXM, 0) {
> + Return (PXM)
> + }
> }
> }
next prev parent reply other threads:[~2013-11-27 15:53 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-11-27 13:02 [Qemu-devel] [RFC PATCH v2] i386: Add _PXM ACPI method to CPU objects Vasilis Liaskovitis
2013-11-27 15:53 ` Igor Mammedov [this message]
2013-11-27 15:58 ` Paolo Bonzini
2013-11-29 14:44 ` Vasilis Liaskovitis
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=20131127165330.075dcf41@nial.usersys.redhat.com \
--to=imammedo@redhat.com \
--cc=mst@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=seabios@seabios.org \
--cc=thilo.fromm@profitbricks.com \
--cc=vasilis.liaskovitis@profitbricks.com \
/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).