From: Lai Jiangshan <jiangshanlai@gmail.com>
To: linux-kernel@vger.kernel.org
Cc: x86@kernel.org, Lai Jiangshan <laijs@linux.alibaba.com>,
Jan Beulich <jbeulich@suse.com>,
Thomas Gleixner <tglx@linutronix.de>,
Juergen Gross <jgross@suse.com>, Peter Anvin <hpa@zytor.com>,
xen-devel@lists.xenproject.org,
Boris Ostrovsky <boris.ostrovsky@oracle.com>,
Andy Lutomirski <luto@kernel.org>, Ingo Molnar <mingo@redhat.com>,
Borislav Petkov <bp@alien8.de>,
Stefano Stabellini <sstabellini@kernel.org>
Subject: [PATCH V3 04/49] x86/xen: Add xenpv_restore_regs_and_return_to_usermode()
Date: Thu, 14 Oct 2021 11:13:27 +0800 [thread overview]
Message-ID: <20211014031413.14471-5-jiangshanlai@gmail.com> (raw)
In-Reply-To: <20211014031413.14471-1-jiangshanlai@gmail.com>
From: Lai Jiangshan <laijs@linux.alibaba.com>
While in the native case, PER_CPU_VAR(cpu_tss_rw + TSS_sp0) is the
trampoline stack. But XEN pv doesn't use trampoline stack, so
PER_CPU_VAR(cpu_tss_rw + TSS_sp0) is also the kernel stack. Hence source
and destination stacks are identical in that case, which means reusing
swapgs_restore_regs_and_return_to_usermode() in XEN pv would cause %rsp
to move up to the top of the kernel stack and leave the IRET frame below
%rsp, which is dangerous to be corrupted if #NMI / #MC hit as either of
these events occurring in the middle of the stack pushing would clobber
data on the (original) stack.
And swapgs_restore_regs_and_return_to_usermode() pushing the IRET frame
on to the original address is useless and error-prone when there is any
future attempt to modify the code.
Fixes: 7f2590a110b8 ("x86/entry/64: Use a per-CPU trampoline stack for IDT entries")
Cc: Jan Beulich <jbeulich@suse.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Juergen Gross <jgross@suse.com>
Cc: Peter Anvin <hpa@zytor.com>
Cc: xen-devel@lists.xenproject.org>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
---
arch/x86/entry/entry_64.S | 9 ++++++---
arch/x86/entry/entry_64_compat.S | 7 ++++---
arch/x86/xen/xen-asm.S | 27 +++++++++++++++++++++++++++
3 files changed, 37 insertions(+), 6 deletions(-)
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index 9d468c8877e2..0dde5a253dda 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -119,7 +119,7 @@ SYM_INNER_LABEL(entry_SYSCALL_64_after_hwframe, SYM_L_GLOBAL)
* In the Xen PV case we must use iret anyway.
*/
- ALTERNATIVE "", "jmp swapgs_restore_regs_and_return_to_usermode", \
+ ALTERNATIVE "", "jmp xenpv_restore_regs_and_return_to_usermode", \
X86_FEATURE_XENPV
movq RCX(%rsp), %rcx
@@ -286,7 +286,8 @@ SYM_CODE_START(ret_from_fork)
UNWIND_HINT_REGS
movq %rsp, %rdi
call syscall_exit_to_user_mode /* returns with IRQs disabled */
- jmp swapgs_restore_regs_and_return_to_usermode
+ ALTERNATIVE "jmp swapgs_restore_regs_and_return_to_usermode", \
+ "jmp xenpv_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
1:
/* kernel thread */
@@ -566,6 +567,7 @@ __irqentry_text_start:
__irqentry_text_end:
SYM_CODE_START_LOCAL(common_interrupt_return)
+SYM_INNER_LABEL(xenpv_restore_regs_and_return_to_usermode, SYM_L_WEAK) /* placeholder */
SYM_INNER_LABEL(swapgs_restore_regs_and_return_to_usermode, SYM_L_GLOBAL)
#ifdef CONFIG_DEBUG_ENTRY
/* Assert that pt_regs indicates user mode. */
@@ -1055,7 +1057,8 @@ SYM_CODE_START_LOCAL(error_return)
DEBUG_ENTRY_ASSERT_IRQS_OFF
testb $3, CS(%rsp)
jz restore_regs_and_return_to_kernel
- jmp swapgs_restore_regs_and_return_to_usermode
+ ALTERNATIVE "jmp swapgs_restore_regs_and_return_to_usermode", \
+ "jmp xenpv_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
SYM_CODE_END(error_return)
/*
diff --git a/arch/x86/entry/entry_64_compat.S b/arch/x86/entry/entry_64_compat.S
index 0051cf5c792d..2a4d9532dfd5 100644
--- a/arch/x86/entry/entry_64_compat.S
+++ b/arch/x86/entry/entry_64_compat.S
@@ -139,7 +139,7 @@ SYM_INNER_LABEL(entry_SYSENTER_compat_after_hwframe, SYM_L_GLOBAL)
call do_SYSENTER_32
/* XEN PV guests always use IRET path */
ALTERNATIVE "testl %eax, %eax; jz swapgs_restore_regs_and_return_to_usermode", \
- "jmp swapgs_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
+ "jmp xenpv_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
jmp sysret32_from_system_call
.Lsysenter_fix_flags:
@@ -256,7 +256,7 @@ SYM_INNER_LABEL(entry_SYSCALL_compat_after_hwframe, SYM_L_GLOBAL)
call do_fast_syscall_32
/* XEN PV guests always use IRET path */
ALTERNATIVE "testl %eax, %eax; jz swapgs_restore_regs_and_return_to_usermode", \
- "jmp swapgs_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
+ "jmp xenpv_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
/* Opportunistic SYSRET */
sysret32_from_system_call:
@@ -411,5 +411,6 @@ SYM_CODE_START(entry_INT80_compat)
movq %rsp, %rdi
call do_int80_syscall_32
- jmp swapgs_restore_regs_and_return_to_usermode
+ ALTERNATIVE "jmp swapgs_restore_regs_and_return_to_usermode", \
+ "jmp xenpv_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
SYM_CODE_END(entry_INT80_compat)
diff --git a/arch/x86/xen/xen-asm.S b/arch/x86/xen/xen-asm.S
index 220dd9678494..032be1bab113 100644
--- a/arch/x86/xen/xen-asm.S
+++ b/arch/x86/xen/xen-asm.S
@@ -20,6 +20,7 @@
#include <linux/init.h>
#include <linux/linkage.h>
+#include <../entry/calling.h>
.pushsection .noinstr.text, "ax"
/*
@@ -192,6 +193,32 @@ SYM_CODE_START(xen_iret)
jmp hypercall_iret
SYM_CODE_END(xen_iret)
+/*
+ * XEN pv doesn't use trampoline stack, PER_CPU_VAR(cpu_tss_rw + TSS_sp0) is
+ * also the kernel stack. Reusing swapgs_restore_regs_and_return_to_usermode()
+ * in XEN pv would cause %rsp to move up to the top of the kernel stack and
+ * leave the IRET frame below %rsp, which is dangerous to be corrupted if #NMI
+ * interrupts. And swapgs_restore_regs_and_return_to_usermode() pushing the IRET
+ * frame at the same address is useless.
+ */
+SYM_CODE_START(xenpv_restore_regs_and_return_to_usermode)
+ UNWIND_HINT_REGS
+#ifdef CONFIG_DEBUG_ENTRY
+ /* Assert that pt_regs indicates user mode. */
+ testb $3, CS(%rsp)
+ jnz 1f
+ ud2
+1:
+#endif
+ POP_REGS
+
+ /* stackleak_erase() can work safely on the kernel stack. */
+ STACKLEAK_ERASE_NOCLOBBER
+
+ addq $8, %rsp /* skip regs->orig_ax */
+ jmp xen_iret
+SYM_CODE_END(xenpv_restore_regs_and_return_to_usermode)
+
/*
* Xen handles syscall callbacks much like ordinary exceptions, which
* means we have:
--
2.19.1.6.gb485710b
next prev parent reply other threads:[~2021-10-14 3:15 UTC|newest]
Thread overview: 57+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-14 3:13 [PATCH V3 00/49] x86/entry/64: Convert a bunch of ASM entry code into C code Lai Jiangshan
2021-10-14 3:13 ` [PATCH V3 01/49] x86/entry: Add fence for kernel entry swapgs in paranoid_entry() Lai Jiangshan
2021-10-14 3:13 ` [PATCH V3 02/49] x86/entry: Use the correct fence macro after swapgs in kernel CR3 Lai Jiangshan
2021-10-14 3:13 ` [PATCH V3 03/49] x86/traps: Remove stack-protector from traps.c Lai Jiangshan
2021-10-14 3:13 ` Lai Jiangshan [this message]
2021-10-16 18:08 ` [PATCH V3 04/49] x86/xen: Add xenpv_restore_regs_and_return_to_usermode() Boris Ostrovsky
2021-10-14 3:13 ` [PATCH V3 05/49] x86/entry: Use swapgs and native_iret directly in swapgs_restore_regs_and_return_to_usermode Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 06/49] compiler_types.h: Add __noinstr_section() for noinstr Lai Jiangshan
2021-10-14 18:29 ` Nick Desaulniers
2021-10-14 18:35 ` Miguel Ojeda
2021-10-18 9:16 ` Lai Jiangshan
2021-10-18 11:29 ` Miguel Ojeda
2021-10-14 3:33 ` [PATCH V3 07/49] x86/entry: Introduce __entry_text for entry code written in C Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 08/49] x86/entry: Move PTI_USER_* to arch/x86/include/asm/processor-flags.h Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 09/49] x86: Remove unused kernel_to_user_p4dp() and user_to_kernel_p4dp() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 10/49] x86: Replace PTI_PGTABLE_SWITCH_BIT with PTI_USER_PGTABLE_BIT Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 11/49] x86: Mark __native_read_cr3() & native_write_cr3() as __always_inline Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 12/49] x86/traps: Move the declaration of native_irq_return_iret into proto.h Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 13/49] x86/entry: Add arch/x86/entry/entry64.c for C entry code Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 14/49] x86/entry: Expose the address of .Lgs_change to entry64.c Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 15/49] x86/entry: Add C verion of SWITCH_TO_KERNEL_CR3 as switch_to_kernel_cr3() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 16/49] x86/traps: Add fence_swapgs_{user,kernel}_entry() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 17/49] x86/entry: Add C {user,kernel}_entry_swapgs_and_fence() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 18/49] x86/traps: Move pt_regs only in fixup_bad_iret() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 19/49] x86/entry: Switch the stack after error_entry() returns Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 20/49] x86/entry: move PUSH_AND_CLEAR_REGS out of error_entry Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 21/49] x86/entry: Move cld to the start of idtentry Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 22/49] x86/entry: Don't call error_entry for XENPV Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 23/49] x86/entry: Convert SWAPGS to swapgs in error_entry() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 24/49] x86/entry: Implement the whole error_entry() as C code Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 25/49] x86/entry: Use idtentry macro for entry_INT80_compat Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 26/49] x86/entry: Convert SWAPGS to swapgs in entry_SYSENTER_compat() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 27/49] x86: Remove the definition of SWAPGS Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 28/49] x86/entry: Make paranoid_exit() callable Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 29/49] x86/entry: Call paranoid_exit() in asm_exc_nmi() Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 30/49] x86/entry: move PUSH_AND_CLEAR_REGS out of paranoid_entry Lai Jiangshan
2021-10-14 3:33 ` [PATCH V3 31/49] x86/entry: Add the C version ist_switch_to_kernel_cr3() Lai Jiangshan
2021-10-14 3:41 ` [PATCH V3 32/49] x86/entry: Add the C version ist_restore_cr3() Lai Jiangshan
2021-10-25 0:34 ` Lai Jiangshan
2021-10-14 3:41 ` [PATCH V3 33/49] x86/entry: Add the C version get_percpu_base() Lai Jiangshan
2021-10-14 3:41 ` [PATCH V3 34/49] x86/entry: Add the C version ist_switch_to_kernel_gsbase() Lai Jiangshan
2021-10-14 3:41 ` [PATCH V3 35/49] x86/entry: Implement the C version ist_paranoid_entry() Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 36/49] x86/entry: Implement the C version ist_paranoid_exit() Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 37/49] x86/entry: Add a C macro to define the function body for IST in .entry.text Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 38/49] x86/debug, mce: Use C entry code Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 39/49] x86/idtentry.h: Move the definitions *IDTENTRY_{MCE|DEBUG}* up Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 40/49] x86/nmi: Use DEFINE_IDTENTRY_NMI for nmi Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 41/49] x86/nmi: Use C entry code Lai Jiangshan
2021-10-14 3:50 ` [PATCH V3 42/49] x86/entry: Add a C macro to define the function body for IST in .entry.text with an error code Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 43/49] x86/doublefault: Use C entry code Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 44/49] x86/sev: Add and use ist_vc_switch_off_ist() Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 45/49] x86/sev: Use C entry code Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 46/49] x86/entry: Remove ASM function paranoid_entry() and paranoid_exit() Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 47/49] x86/entry: Remove the unused ASM macros Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 48/49] x86/entry: Remove save_ret from PUSH_AND_CLEAR_REGS Lai Jiangshan
2021-10-14 3:58 ` [PATCH V3 49/49] x86/syscall/64: Move the checking for sysret to C code Lai Jiangshan
2021-10-23 0:01 ` [PATCH V3 00/49] x86/entry/64: Convert a bunch of ASM entry code into " Lai Jiangshan
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=20211014031413.14471-5-jiangshanlai@gmail.com \
--to=jiangshanlai@gmail.com \
--cc=boris.ostrovsky@oracle.com \
--cc=bp@alien8.de \
--cc=hpa@zytor.com \
--cc=jbeulich@suse.com \
--cc=jgross@suse.com \
--cc=laijs@linux.alibaba.com \
--cc=linux-kernel@vger.kernel.org \
--cc=luto@kernel.org \
--cc=mingo@redhat.com \
--cc=sstabellini@kernel.org \
--cc=tglx@linutronix.de \
--cc=x86@kernel.org \
--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: 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).