From: Boris Ostrovsky <boris.ostrovsky@oracle.com> To: david.vrabel@citrix.com, konrad.wilk@oracle.com Cc: xen-devel@lists.xenproject.org, linux-kernel@vger.kernel.org, roger.pau@citrix.com, mcgrof@suse.com, Boris Ostrovsky <boris.ostrovsky@oracle.com> Subject: [PATCH v2 11/11] xen/hvmlite: Enable CPU on-/offlining Date: Mon, 1 Feb 2016 10:38:57 -0500 [thread overview] Message-ID: <1454341137-14110-12-git-send-email-boris.ostrovsky@oracle.com> (raw) In-Reply-To: <1454341137-14110-1-git-send-email-boris.ostrovsky@oracle.com> When offlining, we should properly clean up interrupts and wait until hypervisor declares VCPU as down before cleaning up. After VCPU that was previously offlined is brought back to life we want to jump back to bare-metal entry points. It's a simple jump on 64-bit but requires minor tweaking for 32-bit case. Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> --- arch/x86/xen/smp.c | 35 +++++++++++++++++++++++++---------- arch/x86/xen/xen-hvmlite.S | 8 ++++++++ 2 files changed, 33 insertions(+), 10 deletions(-) diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index a22cae2..d768c4e 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -143,7 +143,7 @@ static void xen_smp_intr_free(unsigned int cpu) kfree(per_cpu(xen_callfuncsingle_irq, cpu).name); per_cpu(xen_callfuncsingle_irq, cpu).name = NULL; } - if (xen_hvm_domain()) + if (xen_hvm_domain() && !xen_hvmlite) return; if (per_cpu(xen_irq_work, cpu).irq >= 0) { @@ -585,7 +585,8 @@ static int xen_cpu_disable(void) static void xen_cpu_die(unsigned int cpu) { - while (xen_pv_domain() && HYPERVISOR_vcpu_op(VCPUOP_is_up, cpu, NULL)) { + while ((xen_pv_domain() || xen_hvmlite) && + HYPERVISOR_vcpu_op(VCPUOP_is_up, cpu, NULL)) { __set_current_state(TASK_UNINTERRUPTIBLE); schedule_timeout(HZ/10); } @@ -602,14 +603,28 @@ static void xen_play_dead(void) /* used only with HOTPLUG_CPU */ { play_dead_common(); HYPERVISOR_vcpu_op(VCPUOP_down, smp_processor_id(), NULL); - cpu_bringup(); - /* - * commit 4b0c0f294 (tick: Cleanup NOHZ per cpu data on cpu down) - * clears certain data that the cpu_idle loop (which called us - * and that we return from) expects. The only way to get that - * data back is to call: - */ - tick_nohz_idle_enter(); + + if (!xen_hvm_domain()) { + cpu_bringup(); + /* + * commit 4b0c0f294 (tick: Cleanup NOHZ per cpu data on cpu + * down) clears certain data that the cpu_idle loop (which + * called us and that we return from) expects. The only way to + * get that data back is to call: + */ + tick_nohz_idle_enter(); + } else { + /* + * For 64-bit we can jump directly to SMP entry point but for + * 32-bit we need to disable paging and load boot GDT (just + * like in cpu_initialize_context()). + */ +#ifdef CONFIG_X86_64 + asm("jmp secondary_startup_64"); +#else + asm("jmp hvmlite_smp_32_hp"); +#endif + } } #else /* !CONFIG_HOTPLUG_CPU */ diff --git a/arch/x86/xen/xen-hvmlite.S b/arch/x86/xen/xen-hvmlite.S index 805e6a0..1db2e96 100644 --- a/arch/x86/xen/xen-hvmlite.S +++ b/arch/x86/xen/xen-hvmlite.S @@ -145,6 +145,14 @@ ENTRY(hvmlite_start_xen) #endif #ifdef CONFIG_X86_32 +ENTRY(hvmlite_smp_32_hp) + movl $_pa(initial_page_table), %eax + movl %eax, %cr3 + ljmp $__KERNEL_CS,$_pa(5f) +5: + movl $X86_CR0_PE, %eax + movl %eax, %cr0 + ENTRY(hvmlite_smp_32) mov $_pa(boot_gdt_descr), %eax lgdt (%eax) -- 1.7.1
WARNING: multiple messages have this Message-ID (diff)
From: Boris Ostrovsky <boris.ostrovsky@oracle.com> To: david.vrabel@citrix.com, konrad.wilk@oracle.com Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>, xen-devel@lists.xenproject.org, mcgrof@suse.com, linux-kernel@vger.kernel.org, roger.pau@citrix.com Subject: [PATCH v2 11/11] xen/hvmlite: Enable CPU on-/offlining Date: Mon, 1 Feb 2016 10:38:57 -0500 [thread overview] Message-ID: <1454341137-14110-12-git-send-email-boris.ostrovsky@oracle.com> (raw) In-Reply-To: <1454341137-14110-1-git-send-email-boris.ostrovsky@oracle.com> When offlining, we should properly clean up interrupts and wait until hypervisor declares VCPU as down before cleaning up. After VCPU that was previously offlined is brought back to life we want to jump back to bare-metal entry points. It's a simple jump on 64-bit but requires minor tweaking for 32-bit case. Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> --- arch/x86/xen/smp.c | 35 +++++++++++++++++++++++++---------- arch/x86/xen/xen-hvmlite.S | 8 ++++++++ 2 files changed, 33 insertions(+), 10 deletions(-) diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index a22cae2..d768c4e 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -143,7 +143,7 @@ static void xen_smp_intr_free(unsigned int cpu) kfree(per_cpu(xen_callfuncsingle_irq, cpu).name); per_cpu(xen_callfuncsingle_irq, cpu).name = NULL; } - if (xen_hvm_domain()) + if (xen_hvm_domain() && !xen_hvmlite) return; if (per_cpu(xen_irq_work, cpu).irq >= 0) { @@ -585,7 +585,8 @@ static int xen_cpu_disable(void) static void xen_cpu_die(unsigned int cpu) { - while (xen_pv_domain() && HYPERVISOR_vcpu_op(VCPUOP_is_up, cpu, NULL)) { + while ((xen_pv_domain() || xen_hvmlite) && + HYPERVISOR_vcpu_op(VCPUOP_is_up, cpu, NULL)) { __set_current_state(TASK_UNINTERRUPTIBLE); schedule_timeout(HZ/10); } @@ -602,14 +603,28 @@ static void xen_play_dead(void) /* used only with HOTPLUG_CPU */ { play_dead_common(); HYPERVISOR_vcpu_op(VCPUOP_down, smp_processor_id(), NULL); - cpu_bringup(); - /* - * commit 4b0c0f294 (tick: Cleanup NOHZ per cpu data on cpu down) - * clears certain data that the cpu_idle loop (which called us - * and that we return from) expects. The only way to get that - * data back is to call: - */ - tick_nohz_idle_enter(); + + if (!xen_hvm_domain()) { + cpu_bringup(); + /* + * commit 4b0c0f294 (tick: Cleanup NOHZ per cpu data on cpu + * down) clears certain data that the cpu_idle loop (which + * called us and that we return from) expects. The only way to + * get that data back is to call: + */ + tick_nohz_idle_enter(); + } else { + /* + * For 64-bit we can jump directly to SMP entry point but for + * 32-bit we need to disable paging and load boot GDT (just + * like in cpu_initialize_context()). + */ +#ifdef CONFIG_X86_64 + asm("jmp secondary_startup_64"); +#else + asm("jmp hvmlite_smp_32_hp"); +#endif + } } #else /* !CONFIG_HOTPLUG_CPU */ diff --git a/arch/x86/xen/xen-hvmlite.S b/arch/x86/xen/xen-hvmlite.S index 805e6a0..1db2e96 100644 --- a/arch/x86/xen/xen-hvmlite.S +++ b/arch/x86/xen/xen-hvmlite.S @@ -145,6 +145,14 @@ ENTRY(hvmlite_start_xen) #endif #ifdef CONFIG_X86_32 +ENTRY(hvmlite_smp_32_hp) + movl $_pa(initial_page_table), %eax + movl %eax, %cr3 + ljmp $__KERNEL_CS,$_pa(5f) +5: + movl $X86_CR0_PE, %eax + movl %eax, %cr0 + ENTRY(hvmlite_smp_32) mov $_pa(boot_gdt_descr), %eax lgdt (%eax) -- 1.7.1
next prev parent reply other threads:[~2016-02-01 15:45 UTC|newest] Thread overview: 108+ messages / expand[flat|nested] mbox.gz Atom feed top 2016-02-01 15:38 [PATCH v2 00/11] HVMlite domU support Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-01 15:38 ` [PATCH v2 01/11] xen/hvmlite: Import hvmlite-related Xen public interfaces Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-02 16:06 ` David Vrabel 2016-02-02 16:06 ` [Xen-devel] " David Vrabel 2016-02-01 15:38 ` [PATCH v2 02/11] xen/hvmlite: Bootstrap HVMlite guest Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-02 16:39 ` David Vrabel 2016-02-02 16:39 ` [Xen-devel] " David Vrabel 2016-02-02 17:19 ` Boris Ostrovsky 2016-02-02 17:19 ` Boris Ostrovsky 2016-02-03 18:55 ` Luis R. Rodriguez 2016-02-03 18:55 ` Luis R. Rodriguez 2016-02-03 20:11 ` Boris Ostrovsky 2016-02-03 23:40 ` Luis R. Rodriguez 2016-02-03 23:40 ` Luis R. Rodriguez 2016-02-04 19:54 ` Boris Ostrovsky 2016-02-04 20:57 ` Luis R. Rodriguez 2016-02-04 22:28 ` Boris Ostrovsky 2016-02-04 22:28 ` Boris Ostrovsky 2016-02-04 20:57 ` Luis R. Rodriguez 2016-02-04 19:54 ` Boris Ostrovsky 2016-02-03 20:11 ` Boris Ostrovsky 2016-02-03 20:52 ` Andrew Cooper 2016-02-03 20:52 ` [Xen-devel] " Andrew Cooper 2016-02-03 23:59 ` Luis R. Rodriguez 2016-02-04 0:08 ` Luis R. Rodriguez 2016-02-04 0:08 ` [Xen-devel] " Luis R. Rodriguez 2016-02-04 0:51 ` Andrew Cooper 2016-02-04 0:51 ` [Xen-devel] " Andrew Cooper 2016-02-04 23:10 ` Luis R. Rodriguez 2016-04-05 22:02 ` Luis R. Rodriguez 2016-04-05 22:02 ` [Xen-devel] " Luis R. Rodriguez 2016-02-04 23:10 ` Luis R. Rodriguez 2016-02-03 23:59 ` Luis R. Rodriguez 2016-04-24 20:23 ` Borislav Petkov 2016-04-24 20:23 ` Borislav Petkov 2016-04-25 13:21 ` Boris Ostrovsky 2016-04-25 13:21 ` Boris Ostrovsky 2016-04-25 13:47 ` Borislav Petkov 2016-04-25 13:47 ` Borislav Petkov 2016-04-25 13:54 ` Boris Ostrovsky 2016-04-25 13:54 ` Boris Ostrovsky 2016-04-25 14:11 ` Borislav Petkov 2016-04-25 14:11 ` Borislav Petkov 2016-04-25 14:42 ` Boris Ostrovsky 2016-04-25 14:42 ` Boris Ostrovsky 2016-04-25 15:22 ` Borislav Petkov 2016-04-25 15:22 ` Borislav Petkov 2016-04-25 15:48 ` Boris Ostrovsky 2016-04-25 15:48 ` Boris Ostrovsky 2016-04-26 10:53 ` Borislav Petkov 2016-04-26 10:53 ` Borislav Petkov 2016-04-25 17:24 ` David Vrabel 2016-04-25 17:24 ` [Xen-devel] " David Vrabel 2016-02-01 15:38 ` [PATCH v2 03/11] xen/hvmlite: Initialize HVMlite kernel Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-17 20:08 ` [Xen-devel] " Luis R. Rodriguez 2016-02-17 20:08 ` Luis R. Rodriguez 2016-02-01 15:38 ` [PATCH v2 04/11] xen/hvmlite: Allow HVMlite guests delay initializing grant table Boris Ostrovsky 2016-02-02 16:13 ` David Vrabel 2016-02-02 16:13 ` [Xen-devel] " David Vrabel 2016-02-02 16:49 ` Boris Ostrovsky 2016-02-02 16:49 ` Boris Ostrovsky 2016-02-03 18:59 ` Luis R. Rodriguez 2016-02-03 18:59 ` Luis R. Rodriguez 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-01 15:38 ` [PATCH v2 05/11] xen/hvmlite: HVMlite guests always have PV devices Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-01 15:38 ` [PATCH v2 06/11] xen/hvmlite: Prepare cpu_initialize_context() routine for HVMlite SMP Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-02 16:16 ` David Vrabel 2016-02-02 16:16 ` [Xen-devel] " David Vrabel 2016-02-01 15:38 ` [PATCH v2 07/11] xen/hvmlite: Initialize context for secondary VCPUs Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-02 16:21 ` David Vrabel 2016-02-02 16:21 ` [Xen-devel] " David Vrabel 2016-02-02 16:58 ` Boris Ostrovsky 2016-02-02 16:58 ` [Xen-devel] " Boris Ostrovsky 2016-02-04 10:07 ` David Vrabel 2016-02-04 10:07 ` David Vrabel 2016-02-04 12:58 ` Doug Goldstein 2016-02-04 12:58 ` [Xen-devel] " Doug Goldstein 2016-02-04 19:08 ` Boris Ostrovsky 2016-02-04 19:08 ` [Xen-devel] " Boris Ostrovsky 2016-02-01 15:38 ` [PATCH v2 08/11] xen/hvmlite: Extend APIC operations for HVMlite guests Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-04 10:04 ` [Xen-devel] " David Vrabel 2016-02-04 12:14 ` Roger Pau Monné 2016-02-04 12:14 ` [Xen-devel] " Roger Pau Monné 2016-02-04 14:01 ` Boris Ostrovsky 2016-02-04 14:01 ` [Xen-devel] " Boris Ostrovsky 2016-02-04 10:04 ` David Vrabel 2016-02-01 15:38 ` [PATCH v2 09/11] xen/hvmlite: Use x86's default timer init " Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-02 16:27 ` [Xen-devel] " David Vrabel 2016-02-02 17:01 ` Boris Ostrovsky 2016-02-02 17:01 ` [Xen-devel] " Boris Ostrovsky 2016-02-02 16:27 ` David Vrabel 2016-02-01 15:38 ` [PATCH v2 10/11] xen/hvmlite: Boot secondary CPUs Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky 2016-02-04 10:38 ` David Vrabel 2016-02-04 10:38 ` [Xen-devel] " David Vrabel 2016-02-04 13:51 ` Boris Ostrovsky 2016-02-04 13:51 ` [Xen-devel] " Boris Ostrovsky 2016-02-01 15:38 ` Boris Ostrovsky [this message] 2016-02-01 15:38 ` [PATCH v2 11/11] xen/hvmlite: Enable CPU on-/offlining Boris Ostrovsky
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=1454341137-14110-12-git-send-email-boris.ostrovsky@oracle.com \ --to=boris.ostrovsky@oracle.com \ --cc=david.vrabel@citrix.com \ --cc=konrad.wilk@oracle.com \ --cc=linux-kernel@vger.kernel.org \ --cc=mcgrof@suse.com \ --cc=roger.pau@citrix.com \ --cc=xen-devel@lists.xenproject.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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.