From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:40432) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VlhQt-0000su-5q for qemu-devel@nongnu.org; Wed, 27 Nov 2013 10:53:47 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1VlhQp-000301-2n for qemu-devel@nongnu.org; Wed, 27 Nov 2013 10:53:43 -0500 Received: from mx1.redhat.com ([209.132.183.28]:27757) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VlhQo-0002yo-RW for qemu-devel@nongnu.org; Wed, 27 Nov 2013 10:53:39 -0500 Date: Wed, 27 Nov 2013 16:53:30 +0100 From: Igor Mammedov Message-ID: <20131127165330.075dcf41@nial.usersys.redhat.com> In-Reply-To: <1385557363-25756-1-git-send-email-vasilis.liaskovitis@profitbricks.com> References: <1385557363-25756-1-git-send-email-vasilis.liaskovitis@profitbricks.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [RFC PATCH v2] i386: Add _PXM ACPI method to CPU objects List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Vasilis Liaskovitis Cc: thilo.fromm@profitbricks.com, seabios@seabios.org, qemu-devel@nongnu.org, mst@redhat.com On Wed, 27 Nov 2013 14:02:43 +0100 Vasilis Liaskovitis 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 entr= ies. >=20 > Currently, CPU SRAT entries are only enabled for cpus that are already pr= esent > 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 A= CPI > spec mentions "If the Local APIC ID of a dynamically added processor is n= ot > present in the SRAT, a _PXM object must exist for the processor=E2=80=99s= device or one > of its ancestors in the ACPI Namespace." Since SRAT entries are not avail= able > for the hot-pluggable processors, a _PXM method must exist for them. Othe= rwise, > the CPU is hot-added in the wrong NUMA node (default node 0). >=20 > 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, parse= d SRAT > info is reset and lost after hot-remove operations, see kernel commit c4c= 60524. > 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. >=20 > 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 cor= rect > NUMA node, regardless of kernel behaviour. >=20 > v1->v2: > Make method return a DWORD integer > Tested on qemu master + cpu-del patches >=20 > Signed-off-by: Vasilis Liaskovitis > Reviewed-by: Thilo Fromm Patch looks good, Please add patch to update hw/i386/ssdt-proc.hex.generated for hosts withou= t iasl for completness >=20 > --- > hw/i386/acpi-build.c | 5 +++++ > hw/i386/ssdt-proc.dsl | 5 +++++ > 2 files changed, 10 insertions(+) >=20 > 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) > =20 > @@ -724,6 +725,10 @@ build_ssdt(GArray *table_data, GArray *linker, > proc[ACPI_PROC_OFFSET_CPUHEX+1] =3D acpi_get_hex(i); > proc[ACPI_PROC_OFFSET_CPUID1] =3D i; > proc[ACPI_PROC_OFFSET_CPUID2] =3D i; > + proc[ACPI_PROC_OFFSET_CPUPXM] =3D guest_info->node_cpu[i]; > + proc[ACPI_PROC_OFFSET_CPUPXM + 1] =3D 0; > + proc[ACPI_PROC_OFFSET_CPUPXM + 2] =3D 0; > + proc[ACPI_PROC_OFFSET_CPUPXM + 3] =3D 0; > } > =20 > /* 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) > + } > } > }