From: "Jan Beulich" <JBeulich@suse.com>
To: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: xen-devel@lists.xen.org,
Stefano Stabellini <sstabellini@kernel.org>,
Wei Liu <wei.liu2@citrix.com>,
roger.pau@citrix.com
Subject: Re: [PATCH RFC 18/20] libxc/acpi: Build ACPI tables for HVMlite guests
Date: Mon, 06 Jun 2016 07:29:54 -0600 [thread overview]
Message-ID: <575596F202000078000F2098@prv-mh.provo.novell.com> (raw)
In-Reply-To: <1459905949-10358-19-git-send-email-boris.ostrovsky@oracle.com>
>>> On 06.04.16 at 03:25, <boris.ostrovsky@oracle.com> wrote:
> --- /dev/null
> +++ b/tools/libxc/xc_acpi.c
> @@ -0,0 +1,268 @@
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <inttypes.h>
> +#include <assert.h>
> +
> +#include <xen/xen.h>
> +#include <xen/foreign/x86_32.h>
> +#include <xen/foreign/x86_64.h>
Why are these two needed here? The code here shouldn't care
about the mode the guest may later want to run in.
> +#include <xen/hvm/hvm_info_table.h>
> +#include <xen/io/protocols.h>
> +
> +#include "xg_private.h"
> +#include "xc_dom.h"
> +#include "xenctrl.h"
> +
> +#include "acpi2_0.h"
> +
> +#define RESERVED_MEMORY_DYNAMIC_START 0xFC001000
> +#define ACPI_PHYSICAL_ADDRESS 0x000EA020
> +
> +/* Initial allocation for ACPI tables */
> +#define NUM_ACPI_PAGES 16
With which other definitions do these three need to remain in sync?
> +#define PFN(paddr) ((paddr) >> PAGE_SHIFT)
> +
> +extern unsigned char dsdt_anycpu[], dsdt_15cpu[], dsdt_empty[];
const
> +extern int dsdt_anycpu_len, dsdt_15cpu_len, dsdt_empty_len;
And even these may want to be const (and certainly unsigned).
> +static int init_acpi_config(struct xc_dom_image *dom,
> + struct acpi_config *config)
> +{
> + xc_interface *xch = dom->xch;
> + uint32_t domid = dom->guest_domid;
> + xc_dominfo_t info;
> + int i, rc;
> +
> + memset(config, 0, sizeof(*config));
> +
> + config->dsdt_anycpu = config->dsdt_15cpu = dsdt_empty;
> + config->dsdt_anycpu_len = config->dsdt_15cpu_len = dsdt_empty_len;
What good does an empty DSDT do? (Perhaps this question is a
result of there not being any description of this change.)
> + rc = xc_domain_getinfo(xch, domid, 1, &info);
> + if ( rc < 0 )
> + {
> + DOMPRINTF("%s: getdomaininfo failed (rc=%d)", __FUNCTION__, rc);
> + return rc;
> + }
> +
> + config->apic_mode = 1;
> +
> + if ( dom->nr_vnodes )
> + {
> + struct acpi_numa *numa = &config->numa;
> +
> + numa->vmemrange = calloc(dom->nr_vmemranges,
> + sizeof(*numa->vmemrange));
> + numa->vdistance = calloc(dom->nr_vnodes,
> + sizeof(*numa->vdistance));
> + numa->vcpu_to_vnode = calloc(config->nr_vcpus,
> + sizeof(*numa->vcpu_to_vnode));
> + if ( !numa->vmemrange || !numa->vdistance || !numa->vcpu_to_vnode )
> + {
> + DOMPRINTF("%s: Out of memory", __FUNCTION__);
> + free(numa->vmemrange);
> + free(numa->vdistance);
> + free(numa->vcpu_to_vnode);
> + return -ENOMEM;
> + }
> +
> + rc = xc_domain_getvnuma(xch, domid, &numa->nr_vnodes,
> + &numa->nr_vmemranges,
> + &config->nr_vcpus, numa->vmemrange,
> + numa->vdistance, numa->vcpu_to_vnode);
> +
> + if ( rc )
> + {
> + DOMPRINTF("%s: xc_domain_getvnuma failed (rc=%d)", __FUNCTION__, rc);
> + return rc;
> + }
> + }
> + else
> + config->nr_vcpus = info.max_vcpu_id + 1;
> +
> + config->vcpu_online = calloc((HVM_MAX_VCPUS + 7) / 8,
> + sizeof(*config->vcpu_online));
> + if ( config->vcpu_online == NULL )
> + {
> + DOMPRINTF("%s: Can't allocate vcpu_online", __FUNCTION__);
> + return -ENOMEM;
> + }
> +
> + for (i=0; i<config->nr_vcpus; i++)
You seem to be mixing styles here.
> +int xc_dom_build_acpi(struct xc_dom_image *dom)
> +{
> + struct acpi_config config;
> + uint32_t domid = dom->guest_domid;
> + xc_interface *xch = dom->xch;
> + int rc, i, acpi_pages_num;
> + xen_pfn_t extent, *extents;
> + void *acpi_pages, *acpi_physical;
> + void *guest_info_page, *guest_acpi_pages;
> +
> + rc = init_acpi_config(dom, &config);
> + if ( rc )
> + {
> + DOMPRINTF("%s: init_acpi_config failed (rc=%d)", __FUNCTION__, rc);
> + return rc;
> + }
> +
> + /*
> + * Pages to hold ACPI tables and one page for acpi_info, which
> + * will be the first one in this region.
> + */
> + acpi_pages = xc_memalign(xch, PAGE_SIZE, NUM_ACPI_PAGES * PAGE_SIZE);
> + if ( !acpi_pages )
> + {
> + DOMPRINTF("%s: Can't allocate acpi pages", __FUNCTION__);
> + rc = -1;
> + goto out;
> + }
> +
> + config.acpi_info_page = acpi_pages;
> +
> + /* Set up allocator memory */
> + base_addr = alloc_up = (unsigned long)acpi_pages + PAGE_SIZE;
> + alloc_down = (unsigned long)acpi_pages + (NUM_ACPI_PAGES * PAGE_SIZE);
> +
> + /* Map page that will hold RSDP */
> + extent = PFN(ACPI_PHYSICAL_ADDRESS);
> + rc = xc_domain_populate_physmap_exact(xch, domid, 1, 0, 0, &extent);
> + if ( rc )
> + {
> + DOMPRINTF("%s: xc_domain_populate_physmap failed with %d",
> + __FUNCTION__, rc);
> + goto out;
> + }
> + acpi_physical = xc_map_foreign_range(xch, domid, PAGE_SIZE,
> + PROT_READ | PROT_WRITE,
> + PFN(ACPI_PHYSICAL_ADDRESS));
> + if ( !acpi_physical )
> + {
> + DOMPRINTF("%s: Can't map acpi_physical", __FUNCTION__);
> + rc = -1;
> + goto out;
> + }
> +
> + /* Build the tables */
> + acpi_build_tables(&config, (unsigned long)acpi_physical);
> +
> + /* Copy acpi_info page into guest's memory */
> + extent = PFN(ACPI_INFO_PHYSICAL_ADDRESS);
> + rc = xc_domain_populate_physmap_exact(xch, domid, 1, 0, 0, &extent);
> + if ( rc )
> + {
> + DOMPRINTF("%s: xc_domain_populate_physmap failed with %d\n",
> + __FUNCTION__, rc);
> + goto out;
> + }
> + guest_info_page = xc_map_foreign_range(xch, domid, PAGE_SIZE,
> + PROT_READ | PROT_WRITE,
> + PFN(ACPI_INFO_PHYSICAL_ADDRESS));
> + if ( !guest_info_page )
> + {
> + DOMPRINTF("%s: Can't map acpi_info_page", __FUNCTION__);
> + rc = -1;
> + goto out;
> + }
> + memcpy(guest_info_page, acpi_pages, PAGE_SIZE);
> +
> + /* Copy ACPI tables into guest's memory */
> + acpi_pages_num = ((alloc_up - (unsigned long)acpi_pages +
> + (PAGE_SIZE - 1)) >> PAGE_SHIFT) - 1;
This looks like there are acpi_pages_num pages starting at
acpi_pages, yet ...
> + extents = malloc(acpi_pages_num * sizeof(*extents));
> + if ( !extents )
> + {
> + DOMPRINTF("%s: Can't allocate extents array", __FUNCTION__);
> + rc = -ENOMEM;
> + goto out;
> + }
> + for (i = 0; i < acpi_pages_num; i++)
> + extents[i] = PFN(RESERVED_MEMORY_DYNAMIC_START) + i;
> + rc = xc_domain_populate_physmap_exact(xch, domid, acpi_pages_num,
> + 0, 0, extents);
> + if ( rc )
> + {
> + DOMPRINTF("%s: xc_domain_populate_physmap failed with %d",
> + __FUNCTION__, rc);
> + goto out;
> + }
> + guest_acpi_pages = xc_map_foreign_range(xch, domid,
> + PAGE_SIZE * acpi_pages_num,
> + PROT_READ | PROT_WRITE,
> + PFN(RESERVED_MEMORY_DYNAMIC_START));
> + if ( !guest_acpi_pages )
> + {
> + DOMPRINTF("%s Can't map guest_acpi_pages", __FUNCTION__);
> + rc = -1;
> + goto out;
> + }
> +
> + memcpy(guest_acpi_pages, acpi_pages + PAGE_SIZE,
> + acpi_pages_num * PAGE_SIZE);
... this looks like there are acpi_pages_num pages starting at
acpi_pages + PAGE_SIZE.
> --- a/tools/libxc/xc_dom_x86.c
> +++ b/tools/libxc/xc_dom_x86.c
> @@ -643,6 +643,13 @@ static int alloc_magic_pages_hvm(struct xc_dom_image *dom)
> DOMPRINTF("Unable to reserve memory for the start info");
> goto out;
> }
> +
> + rc = xc_dom_build_acpi(dom);
> + if ( rc != 0 )
> + {
> + DOMPRINTF("Unable to build ACPI tables");
> + goto out;
> + }
Iirc there is an "acpi=" guest config setting, yet neither here nor
down the call tree I have been able to find a respective check. Is
that option not relevant anymore? Do we really want to always
have those tables built?
> --- a/tools/libxl/libxl_x86.c
> +++ b/tools/libxl/libxl_x86.c
> @@ -8,15 +8,18 @@ int libxl__arch_domain_prepare_config(libxl__gc *gc,
> xc_domain_configuration_t *xc_config)
> {
>
> - if (d_config->c_info.type == LIBXL_DOMAIN_TYPE_HVM &&
> - d_config->b_info.device_model_version !=
> - LIBXL_DEVICE_MODEL_VERSION_NONE) {
> - /* HVM domains with a device model. */
> - xc_config->emulation_flags = XEN_X86_EMU_ALL;
> - } else {
> - /* PV or HVM domains without a device model. */
> + if (d_config->c_info.type == LIBXL_DOMAIN_TYPE_HVM) {
> + if (d_config->b_info.device_model_version !=
> + LIBXL_DEVICE_MODEL_VERSION_NONE)
> + xc_config->emulation_flags = XEN_X86_EMU_ALL;
> + else
> + /*
> + * HVM guests without device model will want
> + * to have LAPIC emulation.
> + */
> + xc_config->emulation_flags = XEN_X86_EMU_LAPIC;
> + } else
> xc_config->emulation_flags = 0;
> - }
I'm not maintainer of this code, but I think this would read better if
you used an if/else if/else sequence instead of nested if.
> --- a/xen/common/libacpi/build.c
> +++ b/xen/common/libacpi/build.c
> @@ -480,7 +480,7 @@ static int new_vm_gid(struct acpi_config *config)
> return 1;
> }
>
> -void acpi_build_tables(struct acpi_config *config, unsigned int physical)
> +void acpi_build_tables(struct acpi_config *config, unsigned long physical)
I'm having some difficulty seeing how this change belongs here.
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
next prev parent reply other threads:[~2016-06-06 13:29 UTC|newest]
Thread overview: 91+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-04-06 1:25 [PATCH RFC 00/20] Make ACPI builder available to components other than hvmloader Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 01/20] hvmloader: Provide hvmloader_acpi_build_tables() Boris Ostrovsky
2016-06-02 12:42 ` Jan Beulich
2016-06-02 16:39 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 02/20] acpi/hvmloader: Move acpi_info initialization out of ACPI code Boris Ostrovsky
2016-06-02 12:54 ` Jan Beulich
2016-06-02 16:54 ` Boris Ostrovsky
2016-06-03 12:13 ` Jan Beulich
2016-06-03 14:42 ` Boris Ostrovsky
2016-06-03 14:55 ` Jan Beulich
2016-06-03 15:29 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 03/20] acpi/hvmloader: Initialize vm_gid data outside " Boris Ostrovsky
2016-06-02 13:03 ` Jan Beulich
2016-06-02 17:01 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 04/20] acpi/hvmloader: Decide which SSDTs to build in hvmloader Boris Ostrovsky
2016-06-02 13:07 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 05/20] acpi/hvmloader: Move passthrough initialization from ACPI code Boris Ostrovsky
2016-06-02 13:52 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 06/20] acpi/hvmloader: Collect processor and NUMA info in hvmloader Boris Ostrovsky
2016-06-02 14:05 ` Jan Beulich
2016-06-02 17:18 ` Boris Ostrovsky
2016-06-03 12:16 ` Jan Beulich
2016-06-03 14:49 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 07/20] acpi/hvmloader: Set TIS header address " Boris Ostrovsky
2016-06-02 14:09 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 08/20] acpi/hvmloader: Make providing IOAPIC in MADT optional Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 09/20] acpi/hvmloader: Build WAET optionally Boris Ostrovsky
2016-06-02 14:32 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 10/20] acpi/hvmloader: Provide address of acpi_info as an argument to ACPI code Boris Ostrovsky
2016-06-03 16:03 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 11/20] acpi/hvmloader: Translate all addresses when assigning addresses in ACPI tables Boris Ostrovsky
2016-06-06 10:48 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 12/20] acpi/hvmloader: Link ACPI object files directly Boris Ostrovsky
2016-06-06 11:04 ` Jan Beulich
2016-06-06 14:20 ` Boris Ostrovsky
2016-06-06 14:29 ` Jan Beulich
2016-06-06 14:49 ` Boris Ostrovsky
2016-06-06 14:57 ` Jan Beulich
2016-06-06 15:31 ` Andrew Cooper
2016-06-06 15:41 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 13/20] acpi/hvmloader: Add stdio.h, string.h and x86.h Boris Ostrovsky
2016-06-06 11:31 ` Jan Beulich
2016-06-06 15:08 ` Boris Ostrovsky
2016-06-06 15:22 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 14/20] acpi/hvmloader: Replace mem_alloc() and virt_to_phys() with memory ops Boris Ostrovsky
2016-06-06 12:58 ` Jan Beulich
2016-06-06 15:46 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 15/20] acpi: Move ACPI code to xen/common/libacpi Boris Ostrovsky
2016-06-06 13:05 ` Jan Beulich
2016-06-06 16:09 ` Boris Ostrovsky
2016-06-07 6:20 ` Jan Beulich
2016-06-07 12:24 ` Roger Pau Monné
2016-06-07 14:32 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 16/20] x86/vlapic: Don't try to accept 8259 interrupt if !has_vpic() Boris Ostrovsky
2016-06-03 16:14 ` Jan Beulich
2016-06-03 17:50 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 17/20] x86: Allow LAPIC-only emulation_flags for HVM guests Boris Ostrovsky
2016-06-03 16:18 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 18/20] libxc/acpi: Build ACPI tables for HVMlite guests Boris Ostrovsky
2016-06-02 16:26 ` Roger Pau Monné
2016-06-06 12:03 ` Wei Liu
2016-06-06 15:15 ` Boris Ostrovsky
2016-06-16 8:54 ` Wei Liu
2016-06-16 13:07 ` Boris Ostrovsky
2016-06-06 13:29 ` Jan Beulich [this message]
2016-06-06 16:59 ` Boris Ostrovsky
2016-06-07 6:17 ` Jan Beulich
2016-06-07 13:59 ` Boris Ostrovsky
2016-06-07 14:10 ` Jan Beulich
2016-06-07 14:47 ` Boris Ostrovsky
2016-06-07 15:00 ` Jan Beulich
2016-04-06 1:25 ` [PATCH RFC 19/20] acpi: Set HW_REDUCED_ACPI in FADT if IOAPIC is not supported Boris Ostrovsky
2016-06-06 13:38 ` Jan Beulich
2016-06-06 17:31 ` Boris Ostrovsky
2016-06-07 6:06 ` Jan Beulich
2016-06-07 14:02 ` Boris Ostrovsky
2016-06-07 14:12 ` Jan Beulich
2016-06-07 15:17 ` Boris Ostrovsky
2016-06-07 15:41 ` Jan Beulich
2016-06-08 22:04 ` Boris Ostrovsky
2016-06-09 8:13 ` Roger Pau Monné
2016-06-09 8:41 ` Jan Beulich
2016-06-09 14:09 ` Boris Ostrovsky
2016-04-06 1:25 ` [PATCH RFC 20/20] acpi: Make ACPI builder available to hypervisor code Boris Ostrovsky
2016-06-06 13:48 ` Jan Beulich
2016-05-09 20:10 ` [PATCH RFC 00/20] Make ACPI builder available to components other than hvmloader Boris Ostrovsky
2016-05-10 7:25 ` Jan Beulich
2016-06-02 12:40 ` Jan Beulich
2016-06-02 16:37 ` Boris Ostrovsky
2016-06-03 7:18 ` Roger Pau Monné
2016-06-03 10:08 ` Jan Beulich
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=575596F202000078000F2098@prv-mh.provo.novell.com \
--to=jbeulich@suse.com \
--cc=boris.ostrovsky@oracle.com \
--cc=roger.pau@citrix.com \
--cc=sstabellini@kernel.org \
--cc=wei.liu2@citrix.com \
--cc=xen-devel@lists.xen.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).