From: George Dunlap <george.dunlap@eu.citrix.com>
To: xen-devel@lists.xen.org, "Nakajima, Jun" <jun.nakajima@intel.com>,
"Dong, Eddie" <eddie.dong@intel.com>
Subject: Re: [PATCH v14 11/17] pvh: Set up more PV stuff in set_info_guest
Date: Mon, 4 Nov 2013 16:20:12 +0000 [thread overview]
Message-ID: <5277C93C.2060604@eu.citrix.com> (raw)
In-Reply-To: <1383567306-6636-12-git-send-email-george.dunlap@eu.citrix.com>
On 04/11/13 12:15, George Dunlap wrote:
> Allow the guest to set up a few more things when bringing up a vcpu.
>
> This includes cr3 and gs_base.
>
> Also set up wallclock, and only initialize a vcpu once.
>
> Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
> Signed-off-by: Mukesh Rathor <mukesh.rathor@oracle.com>
> ---
> v14:
> - Share more of the codepath, removing a potential bug that might happen
> if paging functions are called with "is_initialised" set to zero.
> - Put cr3 in v->arch.guest_table, so the ref counting happens properly.
> This should fix the "zombie domains" problem.
> v13:
> - Get rid of separate pvh call, and fold gs_base write into hvm_set_info_guest
> - Check pvh parameters for validity at the top of arch_set_info_guest
> - Fix comment about PVH and set_info_guest
>
> CC: Jan Beulich <jbeulich@suse.com>
> CC: Tim Deegan <tim@xen.org>
> CC: Keir Fraser <keir@xen.org>
> ---
> xen/arch/x86/domain.c | 30 ++++++++++++++++++++++++++++--
> xen/arch/x86/hvm/vmx/vmx.c | 7 ++++++-
> xen/include/asm-x86/hvm/hvm.h | 6 +++---
> xen/include/public/arch-x86/xen.h | 11 +++++++++++
> 4 files changed, 48 insertions(+), 6 deletions(-)
>
> diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
> index 8c2a57f..c80ef4c 100644
> --- a/xen/arch/x86/domain.c
> +++ b/xen/arch/x86/domain.c
> @@ -691,6 +691,18 @@ int arch_set_info_guest(
> (c(ldt_ents) > 8192) )
> return -EINVAL;
> }
> + else if ( is_pvh_vcpu(v) )
> + {
> + /* PVH 32bitfixme */
> + ASSERT(!compat);
> +
> + if ( c(ctrlreg[1]) || c(ldt_base) || c(ldt_ents) ||
> + c(user_regs.cs) || c(user_regs.ss) || c(user_regs.es) ||
> + c(user_regs.ds) || c(user_regs.fs) || c(user_regs.gs) ||
> + c.nat->gdt_ents || c.nat->fs_base || c.nat->gs_base_user )
> + return -EINVAL;
> +
> + }
>
> v->fpu_initialised = !!(flags & VGCF_I387_VALID);
>
> @@ -728,8 +740,21 @@ int arch_set_info_guest(
>
> if ( has_hvm_container_vcpu(v) )
> {
> - hvm_set_info_guest(v);
> - goto out;
> + hvm_set_info_guest(v, compat ? 0 : c.nat->gs_base_kernel);
> +
> + if ( is_hvm_vcpu(v) || v->is_initialised )
> + goto out;
> +
> + cr3_gfn = xen_cr3_to_pfn(c.nat->ctrlreg[3]);
> + cr3_page = get_page_from_gfn(d, cr3_gfn, NULL, P2M_ALLOC);
> +
> + v->arch.cr3 = page_to_maddr(cr3_page);
> + v->arch.hvm_vcpu.guest_cr[3] = c.nat->ctrlreg[3];
> + v->arch.guest_table = pagetable_from_page(cr3_page);
> +
> + ASSERT(paging_mode_enabled(d));
> +
> + goto pvh_skip_pv_stuff;
> }
>
> init_int80_direct_trap(v);
> @@ -934,6 +959,7 @@ int arch_set_info_guest(
>
> clear_bit(_VPF_in_reset, &v->pause_flags);
>
> + pvh_skip_pv_stuff:
> if ( v->vcpu_id == 0 )
> update_domain_wallclock_time(d);
>
> diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
> index fdb560e..94e9e21 100644
> --- a/xen/arch/x86/hvm/vmx/vmx.c
> +++ b/xen/arch/x86/hvm/vmx/vmx.c
> @@ -1401,7 +1401,7 @@ static void vmx_set_uc_mode(struct vcpu *v)
> hvm_asid_flush_vcpu(v);
> }
>
> -static void vmx_set_info_guest(struct vcpu *v)
> +static void vmx_set_info_guest(struct vcpu *v, uint64_t gs_base_kernel)
> {
> unsigned long intr_shadow;
>
> @@ -1426,6 +1426,11 @@ static void vmx_set_info_guest(struct vcpu *v)
> __vmwrite(GUEST_INTERRUPTIBILITY_INFO, intr_shadow);
> }
>
> + /* PVH 32bitfixme */
> + if ( is_pvh_vcpu(v) )
> + __vmwrite(GUEST_GS_BASE, gs_base_kernel);
> +
> +
> vmx_vmcs_exit(v);
> }
>
> diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
> index 3376418..d6bfcf2 100644
> --- a/xen/include/asm-x86/hvm/hvm.h
> +++ b/xen/include/asm-x86/hvm/hvm.h
> @@ -157,7 +157,7 @@ struct hvm_function_table {
> int (*msr_write_intercept)(unsigned int msr, uint64_t msr_content);
> void (*invlpg_intercept)(unsigned long vaddr);
> void (*set_uc_mode)(struct vcpu *v);
> - void (*set_info_guest)(struct vcpu *v);
> + void (*set_info_guest)(struct vcpu *v, uint64_t gs_base_kernel);
> void (*set_rdtsc_exiting)(struct vcpu *v, bool_t);
>
> /* Nested HVM */
> @@ -431,10 +431,10 @@ void *hvm_map_guest_frame_rw(unsigned long gfn, bool_t permanent);
> void *hvm_map_guest_frame_ro(unsigned long gfn, bool_t permanent);
> void hvm_unmap_guest_frame(void *p, bool_t permanent);
>
> -static inline void hvm_set_info_guest(struct vcpu *v)
> +static inline void hvm_set_info_guest(struct vcpu *v, uint64_t gs_base_kernel)
> {
> if ( hvm_funcs.set_info_guest )
> - return hvm_funcs.set_info_guest(v);
> + return hvm_funcs.set_info_guest(v, gs_base_kernel);
> }
>
> int hvm_debug_op(struct vcpu *v, int32_t op);
> diff --git a/xen/include/public/arch-x86/xen.h b/xen/include/public/arch-x86/xen.h
> index 908ef87..42b818e 100644
> --- a/xen/include/public/arch-x86/xen.h
> +++ b/xen/include/public/arch-x86/xen.h
> @@ -154,6 +154,17 @@ typedef uint64_t tsc_timestamp_t; /* RDTSC timestamp */
> /*
> * The following is all CPU context. Note that the fpu_ctxt block is filled
> * in by FXSAVE if the CPU has feature FXSR; otherwise FSAVE is used.
> + *
> + * Also note that when calling DOMCTL_setvcpucontext and VCPU_initialise
> + * for HVM and PVH guests, not all information in this structure is updated:
> + *
> + * - For HVM guests, the structures read include: fpu_ctxt (if
> + * VGCT_I387_VALID is set), flags, user_regs, debugreg[*]
> + *
> + * - PVH guests are the same as HVM guests, but additionally set cr3,
> + * and for 64-bit guests, gs_base_kernel. Additionally, the following
> + * entries must be 0: ctrlreg[1], ldt_base, ldg_ents, user_regs.{cs,
> + * ss, es, ds, fs, gs), gdt_ents, fs_base, and gs_base_user.
> */
> struct vcpu_guest_context {
> /* FPU registers come first so they can be aligned for FXSAVE/FXRSTOR. */
next prev parent reply other threads:[~2013-11-04 16:20 UTC|newest]
Thread overview: 64+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-11-04 12:14 [PATCH v14 00/20] Introduce PVH domU support George Dunlap
2013-11-04 12:14 ` [PATCH v14 01/17] Allow vmx_update_debug_state to be called when v!=current George Dunlap
2013-11-04 16:01 ` Jan Beulich
2013-11-04 16:18 ` George Dunlap
2013-11-04 12:14 ` [PATCH v14 02/17] libxc: Move temporary grant table mapping to end of memory George Dunlap
2013-11-05 10:57 ` Roger Pau Monné
2013-11-05 11:01 ` Ian Campbell
2013-11-04 12:14 ` [PATCH v14 03/17] pvh prep: code motion George Dunlap
2013-11-04 16:14 ` Jan Beulich
2013-11-07 10:48 ` George Dunlap
2013-11-04 12:14 ` [PATCH v14 04/17] Introduce pv guest type and has_hvm_container macros George Dunlap
2013-11-04 16:20 ` Jan Beulich
2013-11-04 16:26 ` George Dunlap
2013-11-04 16:39 ` George Dunlap
2013-11-07 10:55 ` George Dunlap
2013-11-07 11:04 ` Jan Beulich
2013-11-07 11:11 ` George Dunlap
2013-11-04 12:14 ` [PATCH v14 05/17] pvh: Introduce PVH guest type George Dunlap
2013-11-06 23:28 ` Tim Deegan
2013-11-07 11:21 ` George Dunlap
2013-11-07 16:59 ` Tim Deegan
2013-11-04 12:14 ` [PATCH v14 06/17] pvh: Disable unneeded features of HVM containers George Dunlap
2013-11-04 16:21 ` George Dunlap
2013-11-04 16:37 ` Jan Beulich
2013-11-06 23:54 ` Tim Deegan
2013-11-07 9:00 ` Jan Beulich
2013-11-07 17:02 ` Tim Deegan
2013-11-04 12:14 ` [PATCH v14 07/17] pvh: vmx-specific changes George Dunlap
2013-11-04 16:19 ` George Dunlap
2013-11-04 16:42 ` Jan Beulich
2013-11-07 0:28 ` Tim Deegan
2013-11-07 0:27 ` Tim Deegan
2013-11-07 14:50 ` George Dunlap
2013-11-07 15:40 ` Andrew Cooper
2013-11-07 15:43 ` George Dunlap
2013-11-07 17:00 ` Tim Deegan
2013-11-04 12:14 ` [PATCH v14 08/17] pvh: Do not allow PVH guests to change paging modes George Dunlap
2013-11-04 12:14 ` [PATCH v14 09/17] pvh: PVH access to hypercalls George Dunlap
2013-11-04 12:14 ` [PATCH v14 10/17] pvh: Use PV e820 George Dunlap
2013-11-04 12:15 ` [PATCH v14 11/17] pvh: Set up more PV stuff in set_info_guest George Dunlap
2013-11-04 16:20 ` George Dunlap [this message]
2013-11-04 16:53 ` Jan Beulich
2013-11-07 15:51 ` George Dunlap
2013-11-07 16:10 ` Jan Beulich
2013-11-07 16:33 ` George Dunlap
2013-11-04 12:15 ` [PATCH v14 12/17] pvh: Use PV handlers for cpuid, and IO George Dunlap
2013-11-04 16:20 ` George Dunlap
2013-11-05 8:42 ` Jan Beulich
2013-11-07 16:50 ` George Dunlap
2013-11-04 12:15 ` [PATCH v14 13/17] pvh: Disable 32-bit guest support for now George Dunlap
2013-11-04 12:15 ` [PATCH v14 14/17] pvh: Restrict tsc_mode to NEVER_EMULATE " George Dunlap
2013-11-04 12:15 ` [PATCH v14 15/17] pvh: Documentation George Dunlap
2013-11-04 12:15 ` [PATCH v14 16/17] PVH xen tools: libxc changes to build a PVH guest George Dunlap
2013-11-04 12:15 ` [PATCH v14 17/17] PVH xen tools: libxl changes to create " George Dunlap
2013-11-04 16:59 ` [PATCH v14 00/20] Introduce PVH domU support Konrad Rzeszutek Wilk
2013-11-04 17:23 ` George Dunlap
2013-11-04 17:34 ` Tim Deegan
2013-11-08 15:41 ` George Dunlap
2013-11-08 15:53 ` George Dunlap
2013-11-08 17:01 ` Tim Deegan
2013-11-08 17:06 ` George Dunlap
2013-11-08 15:58 ` Konrad Rzeszutek Wilk
2013-11-07 1:11 ` Tim Deegan
2013-11-11 12:37 ` Roger Pau Monné
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=5277C93C.2060604@eu.citrix.com \
--to=george.dunlap@eu.citrix.com \
--cc=eddie.dong@intel.com \
--cc=jun.nakajima@intel.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).