Skip to contentSkip to conversation

x86/entry: Fixes

PATCH Applied Andy Lutomirski 6 participants 18 messages Jun 26, 2020 – Jul 6, 2020

Subsystems (4)

v1 of 2 Applied Latest is v2 →

lore
Share & tools
lore

Andy Lutomirski

author maintainer
#0 lore raw

cover x86/entry: Fixes

Half of this is selftests. The rest adds some confidence-improving assertions and fixes a crash when running on Xen PV.

Andy Lutomirski (6):
x86/entry: Assert that syscalls are on the right stack x86/entry: Move SYSENTER's regs->sp and regs->flags fixups into C x86/entry/64/compat: Fix Xen PV SYSENTER frame setup selftests/x86/syscall_nt: Add more flag combinations selftests/x86/syscall_nt: Clear weird flags after each test selftests/x86: Consolidate and fix get/set_eflags() helpers

 arch/x86/entry/common.c                       | 30 ++++++++++++--
 arch/x86/entry/entry_32.S                     |  5 +--
 arch/x86/entry/entry_64_compat.S              | 12 +++---
 arch/x86/xen/xen-asm_64.S                     | 20 +++++++--
 tools/testing/selftests/x86/Makefile          |  4 +-
 tools/testing/selftests/x86/helpers.h         | 41 +++++++++++++++++++
 .../selftests/x86/single_step_syscall.c       | 17 +-------
 .../testing/selftests/x86/syscall_arg_fault.c | 21 +---------
 tools/testing/selftests/x86/syscall_nt.c      | 36 ++++++++--------
 tools/testing/selftests/x86/test_vsyscall.c   | 15 +------
 tools/testing/selftests/x86/unwind_vdso.c     | 23 +----------
 11 files changed, 118 insertions(+), 106 deletions(-)
 create mode 100644 tools/testing/selftests/x86/helpers.h
Signature · 1 line
2.25.4
author maintainer
#1 lore raw

PATCH 1/6 x86/entry: Assert that syscalls are on the right stack

Now that the entry stack is a full page, it's too easy to regress the system call entry code and end up on the wrong stack without noticing. Assert that all system calls (SYSCALL64, SYSCALL32, SYSENTER, and INT80) are on the right stack and have pt_regs in the right place.

---
 arch/x86/entry/common.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)
arch/x86/entry/common.c +15 -3
index bd3f14175193..ed8ccc820995 100644
@@ -45,6 +45,15 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/syscalls.h>
 
+/* Check that the stack and regs on entry from user mode are sane. */
+static void check_user_regs(struct pt_regs *regs)
+{
+	if (IS_ENABLED(CONFIG_DEBUG_ENTRY)) {
+		WARN_ON_ONCE(!on_thread_stack());
+		WARN_ON_ONCE(regs != task_pt_regs(current));
+	}
+}
+
 #ifdef CONFIG_CONTEXT_TRACKING
 /**
  * enter_from_user_mode - Establish state when coming from user mode
@@ -127,9 +136,6 @@ static long syscall_trace_enter(struct pt_regs *regs)
 	unsigned long ret = 0;
 	u32 work;
 
-	if (IS_ENABLED(CONFIG_DEBUG_ENTRY))
-		BUG_ON(regs != task_pt_regs(current));
-
 	work = READ_ONCE(ti->flags);
 
 	if (work & (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_EMU)) {
@@ -346,6 +352,8 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs)
 {
 	struct thread_info *ti;
 
+	check_user_regs(regs);
+
 	enter_from_user_mode();
 	instrumentation_begin();
 
@@ -409,6 +417,8 @@ static void do_syscall_32_irqs_on(struct pt_regs *regs)
 /* Handles int $0x80 */
 __visible noinstr void do_int80_syscall_32(struct pt_regs *regs)
 {
+	check_user_regs(regs);
+
 	enter_from_user_mode();
 	instrumentation_begin();
 
@@ -460,6 +470,8 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)
 					vdso_image_32.sym_int80_landing_pad;
 	bool success;
 
+	check_user_regs(regs);
+
 	/*
 	 * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
 	 * so that 'regs->ip -= 2' lands back on an int $0x80 instruction.
Signature · 1 line
2.25.4
author maintainer
#14 lore raw

PATCH 5/6 selftests/x86/syscall_nt: Clear weird flags after each test

In reply to Andy Lutomirski

Clear the weird flags before logging to improve strace output -- logging results while, say, TF is set does no one any favors.

---
 tools/testing/selftests/x86/syscall_nt.c | 1 +
 1 file changed, 1 insertion(+)
tools/testing/selftests/x86/syscall_nt.c +1 -0
index f060534b66a0..5fc82b9cebed 100644
@@ -59,6 +59,7 @@ static void do_it(unsigned long extraflags)
 	set_eflags(get_eflags() | extraflags);
 	syscall(SYS_getpid);
 	flags = get_eflags();
+	set_eflags(X86_EFLAGS_IF | X86_EFLAGS_FIXED);
 	if ((flags & extraflags) == extraflags) {
 		printf("[OK]\tThe syscall worked and flags are still set\n");
 	} else {
Signature · 1 line
2.25.4
author maintainer
#12 lore raw

PATCH 4/6 selftests/x86/syscall_nt: Add more flag combinations

In reply to Andy Lutomirski

Add EFLAGS.AC to the mix.

---
 tools/testing/selftests/x86/syscall_nt.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
tools/testing/selftests/x86/syscall_nt.c +15 -0
index 02309a195041..f060534b66a0 100644
@@ -73,6 +73,12 @@ int main(void)
 	printf("[RUN]\tSet NT and issue a syscall\n");
 	do_it(X86_EFLAGS_NT);
 
+	printf("[RUN]\tSet AC and issue a syscall\n");
+	do_it(X86_EFLAGS_AC);
+
+	printf("[RUN]\tSet NT|AC and issue a syscall\n");
+	do_it(X86_EFLAGS_NT | X86_EFLAGS_AC);
+
 	/*
 	 * Now try it again with TF set -- TF forces returns via IRET in all
 	 * cases except non-ptregs-using 64-bit full fast path syscalls.
@@ -80,8 +86,17 @@ int main(void)
 
 	sethandler(SIGTRAP, sigtrap, 0);
 
+	printf("[RUN]\tSet TF and issue a syscall\n");
+	do_it(X86_EFLAGS_TF);
+
 	printf("[RUN]\tSet NT|TF and issue a syscall\n");
 	do_it(X86_EFLAGS_NT | X86_EFLAGS_TF);
 
+	printf("[RUN]\tSet AC|TF and issue a syscall\n");
+	do_it(X86_EFLAGS_AC | X86_EFLAGS_TF);
+
+	printf("[RUN]\tSet NT|AC|TF and issue a syscall\n");
+	do_it(X86_EFLAGS_NT | X86_EFLAGS_AC | X86_EFLAGS_TF);
+
 	return nerrs == 0 ? 0 : 1;
 }
Signature · 1 line
2.25.4
author maintainer
#6 lore raw

PATCH 3/6 x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

In reply to Andy Lutomirski

The SYSENTER frame setup was nonsense. It worked by accident because the normal code into which the Xen asm jumped (entry_SYSENTER_32/compat) threw away SP without touching the stack. entry_SYSENTER_compat was recently modified such that it relied on having a valid stack pointer, so now the Xen asm needs to invoke it with a valid stack.

Fix it up like SYSCALL: use the Xen-provided frame and skip the bare metal prologue.

Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Fixes 1c3e5d3f60e2 ("x86/entry: Make entry_64_compat.S objtool clean")
---
 arch/x86/entry/entry_64_compat.S |  1 +
 arch/x86/xen/xen-asm_64.S        | 20 ++++++++++++++++----
 2 files changed, 17 insertions(+), 4 deletions(-)
arch/x86/entry/entry_64_compat.S +1 -0
index 7b9d8150f652..381a6de7de9c 100644
@@ -79,6 +79,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
 	pushfq				/* pt_regs->flags (except IF = 0) */
 	pushq	$__USER32_CS		/* pt_regs->cs */
 	pushq	$0			/* pt_regs->ip = 0 (placeholder) */
+SYM_INNER_LABEL(entry_SYSENTER_compat_after_hwframe, SYM_L_GLOBAL)
 	pushq	%rax			/* pt_regs->orig_ax */
 	pushq	%rdi			/* pt_regs->di */
 	pushq	%rsi			/* pt_regs->si */
arch/x86/xen/xen-asm_64.S +16 -4
index 5d252aaeade8..e1e1c7eafa60 100644
@@ -161,10 +161,22 @@ SYM_FUNC_END(xen_syscall32_target)
 
 /* 32-bit compat sysenter target */
 SYM_FUNC_START(xen_sysenter_target)
-	mov 0*8(%rsp), %rcx
-	mov 1*8(%rsp), %r11
-	mov 5*8(%rsp), %rsp
-	jmp entry_SYSENTER_compat
+	/*
+	 * NB: Xen is polite and clears TF from EFLAGS for us.  This means
+	 * that we don't need to guard against single step exceptions here.
+	 */
+	popq %rcx
+	popq %r11
+
+	/*
+	 * Neither Xen nor the kernel really knows what the old SS and
+	 * CS were.  The kernel expects __USER32_DS and __USER32_CS, so
+	 * report those values even though Xen will guess its own values.
+	 */
+	movq $__USER32_DS, 4*8(%rsp)
+	movq $__USER32_CS, 1*8(%rsp)
+
+	jmp entry_SYSENTER_compat_after_hwframe
 SYM_FUNC_END(xen_sysenter_target)
 
 #else /* !CONFIG_IA32_EMULATION */
Signature · 1 line
2.25.4
author maintainer
#16 lore raw

PATCH 6/6 selftests/x86: Consolidate and fix get/set_eflags() helpers

In reply to Andy Lutomirski

We had several copies of get_eflags() and set_eflags() and they were all buggy. Consolidate them and fix them. The fixes are:

Add memory clobbers. These are probably unnecessary but they make sure sure that the compiler doesn't move something past one of these calls when it shouldn't.

Respect the redzone on x86_64. I haven't seen a failure related to this, but it's definitely a bug.

---
 tools/testing/selftests/x86/Makefile          |  4 +-
 tools/testing/selftests/x86/helpers.h         | 41 +++++++++++++++++++
 .../selftests/x86/single_step_syscall.c       | 17 +-------
 .../testing/selftests/x86/syscall_arg_fault.c | 21 +---------
 tools/testing/selftests/x86/syscall_nt.c      | 20 +--------
 tools/testing/selftests/x86/test_vsyscall.c   | 15 +------
 tools/testing/selftests/x86/unwind_vdso.c     | 23 +----------
 7 files changed, 51 insertions(+), 90 deletions(-)
 create mode 100644 tools/testing/selftests/x86/helpers.h
tools/testing/selftests/x86/Makefile +2 -2
index 3ff94575d02d..6703c7906b71 100644
@@ -70,10 +70,10 @@ all_64: $(BINARIES_64)
 
 EXTRA_CLEAN := $(BINARIES_32) $(BINARIES_64)
 
-$(BINARIES_32): $(OUTPUT)/%_32: %.c
+$(BINARIES_32): $(OUTPUT)/%_32: %.c helpers.h
 	$(CC) -m32 -o $@ $(CFLAGS) $(EXTRA_CFLAGS) $^ -lrt -ldl -lm
 
-$(BINARIES_64): $(OUTPUT)/%_64: %.c
+$(BINARIES_64): $(OUTPUT)/%_64: %.c helpers.h
 	$(CC) -m64 -o $@ $(CFLAGS) $(EXTRA_CFLAGS) $^ -lrt -ldl
 
 # x86_64 users should be encouraged to install 32-bit libraries
tools/testing/selftests/x86/helpers.h new file +41 -0

41 lines not on this page · raw message

tools/testing/selftests/x86/single_step_syscall.c +2 -15
index 1063328e275c..120ac741fe44 100644
@@ -31,6 +31,8 @@
 #include <sys/ptrace.h>
 #include <sys/user.h>
 
+#include "helpers.h"
+
 static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		       int flags)
 {
@@ -67,21 +69,6 @@ static unsigned char altstack_data[SIGSTKSZ];
 # define INT80_CLOBBERS
 #endif
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
-#define X86_EFLAGS_TF (1UL << 8)
-
 static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
 {
 	ucontext_t *ctx = (ucontext_t*)ctx_void;
tools/testing/selftests/x86/syscall_arg_fault.c +1 -20
index bc0ecc2e862e..5b7abebbcbb9 100644
@@ -15,30 +15,11 @@
 #include <setjmp.h>
 #include <errno.h>
 
-#ifdef __x86_64__
-# define WIDTH "q"
-#else
-# define WIDTH "l"
-#endif
+#include "helpers.h"
 
 /* Our sigaltstack scratch space. */
 static unsigned char altstack_data[SIGSTKSZ];
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
-#define X86_EFLAGS_TF (1UL << 8)
-
 static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		       int flags)
 {
tools/testing/selftests/x86/syscall_nt.c +1 -19
index 5fc82b9cebed..970e5e14d96d 100644
@@ -13,29 +13,11 @@
 #include <signal.h>
 #include <err.h>
 #include <sys/syscall.h>
-#include <asm/processor-flags.h>
 
-#ifdef __x86_64__
-# define WIDTH "q"
-#else
-# define WIDTH "l"
-#endif
+#include "helpers.h"
 
 static unsigned int nerrs;
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
 static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		       int flags)
 {
tools/testing/selftests/x86/test_vsyscall.c +2 -13
index a4f4d4cf22c3..c41f24b517f4 100644
@@ -20,6 +20,8 @@
 #include <setjmp.h>
 #include <sys/uio.h>
 
+#include "helpers.h"
+
 #ifdef __x86_64__
 # define VSYS(x) (x)
 #else
@@ -493,21 +495,8 @@ static int test_process_vm_readv(void)
 }
 
 #ifdef __x86_64__
-#define X86_EFLAGS_TF (1UL << 8)
 static volatile sig_atomic_t num_vsyscall_traps;
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushfq\n\tpopq %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("pushq %0\n\tpopfq" : : "rm" (eflags) : "flags");
-}
-
 static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
 {
 	ucontext_t *ctx = (ucontext_t *)ctx_void;
tools/testing/selftests/x86/unwind_vdso.c +2 -21
index 0075ccd65407..4c311e1af4c7 100644
@@ -11,6 +11,8 @@
 #include <features.h>
 #include <stdio.h>
 
+#include "helpers.h"
+
 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ < 16
 
 int main()
@@ -53,27 +55,6 @@ static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		err(1, "sigaction");
 }
 
-#ifdef __x86_64__
-# define WIDTH "q"
-#else
-# define WIDTH "l"
-#endif
-
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
-#define X86_EFLAGS_TF (1UL << 8)
-
 static volatile sig_atomic_t nerrs;
 static unsigned long sysinfo;
 static bool got_sysinfo = false;
Signature · 1 line
2.25.4
author maintainer
#4 lore raw

PATCH 2/6 x86/entry: Move SYSENTER's regs->sp and regs->flags fixups into C

In reply to Andy Lutomirski

The SYSENTER asm (32-bit and compat) contains fixups for regs->sp and regs->flags. Move the fixups into C and fix some comments while we're at it.

IMO this is a valid cleanup all by itself, and it also simplifies the subsequent patch that will fix Xen PV SYSENTER.

---
 arch/x86/entry/common.c          | 12 ++++++++++++
 arch/x86/entry/entry_32.S        |  5 ++---
 arch/x86/entry/entry_64_compat.S | 11 +++++------
 3 files changed, 19 insertions(+), 9 deletions(-)
arch/x86/entry/common.c +12 -0
index ed8ccc820995..f392a8bcd1c3 100644
@@ -522,6 +522,18 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)
 		(regs->flags & (X86_EFLAGS_RF | X86_EFLAGS_TF | X86_EFLAGS_VM)) == 0;
 #endif
 }
+
+/* Returns 0 to return using IRET or 1 to return using SYSEXIT/SYSRETL. */
+__visible noinstr long do_SYSENTER_32(struct pt_regs *regs)
+{
+	/* SYSENTER loses RSP, but the vDSO saved it in RBP. */
+	regs->sp = regs->bp;
+
+	/* SYSENTER clobbers EFLAGS.IF.  Assume it was set in usermode. */
+	regs->flags |= X86_EFLAGS_IF;
+
+	return do_fast_syscall_32(regs);
+}
 #endif
 
 SYSCALL_DEFINE0(ni_syscall)
arch/x86/entry/entry_32.S +2 -3
index 024d7d276cd4..2d0bd5d5f032 100644
@@ -933,9 +933,8 @@ SYM_FUNC_START(entry_SYSENTER_32)
 
 .Lsysenter_past_esp:
 	pushl	$__USER_DS		/* pt_regs->ss */
-	pushl	%ebp			/* pt_regs->sp (stashed in bp) */
+	pushl	$0			/* pt_regs->sp (placeholder) */
 	pushfl				/* pt_regs->flags (except IF = 0) */
-	orl	$X86_EFLAGS_IF, (%esp)	/* Fix IF */
 	pushl	$__USER_CS		/* pt_regs->cs */
 	pushl	$0			/* pt_regs->ip = 0 (placeholder) */
 	pushl	%eax			/* pt_regs->orig_ax */
@@ -965,7 +964,7 @@ SYM_FUNC_START(entry_SYSENTER_32)
 .Lsysenter_flags_fixed:
 
 	movl	%esp, %eax
-	call	do_fast_syscall_32
+	call	do_SYSENTER_32
 	/* XEN PV guests always use IRET path */
 	ALTERNATIVE "testl %eax, %eax; jz .Lsyscall_32_done", \
 		    "jmp .Lsyscall_32_done", X86_FEATURE_XENPV
arch/x86/entry/entry_64_compat.S +5 -6
index 0f974ae01e62..7b9d8150f652 100644
@@ -68,16 +68,15 @@ SYM_CODE_START(entry_SYSENTER_compat)
 
 	/* Construct struct pt_regs on stack */
 	pushq	$__USER32_DS		/* pt_regs->ss */
-	pushq	%rbp			/* pt_regs->sp (stashed in bp) */
+	pushq	$0			/* pt_regs->sp = 0 (placeholder) */
 
 	/*
 	 * Push flags.  This is nasty.  First, interrupts are currently
-	 * off, but we need pt_regs->flags to have IF set.  Second, even
-	 * if TF was set when SYSENTER started, it's clear by now.  We fix
-	 * that later using TIF_SINGLESTEP.
+	 * off, but we need pt_regs->flags to have IF set.  Second, if TS
+	 * was set in usermode, it's still set, and we're singlestepping
+	 * through this code.  do_SYSENTER_32() will fix up IF.
 	 */
 	pushfq				/* pt_regs->flags (except IF = 0) */
-	orl	$X86_EFLAGS_IF, (%rsp)	/* Fix saved flags */
 	pushq	$__USER32_CS		/* pt_regs->cs */
 	pushq	$0			/* pt_regs->ip = 0 (placeholder) */
 	pushq	%rax			/* pt_regs->orig_ax */
@@ -135,7 +134,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
 .Lsysenter_flags_fixed:
 
 	movq	%rsp, %rdi
-	call	do_fast_syscall_32
+	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
Signature · 1 line
2.25.4
reviewer Reviewed-by on v1 3/6
#7 lore raw

Re: [PATCH 3/6] x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

In reply to Andy Lutomirski
Quoted 10 lines · Andy Lutomirski

The SYSENTER frame setup was nonsense. It worked by accident because the normal code into which the Xen asm jumped (entry_SYSENTER_32/compat) threw away SP without touching the stack. entry_SYSENTER_compat was recently modified such that it relied on having a valid stack pointer, so now the Xen asm needs to invoke it with a valid stack.

Fix it up like SYSCALL: use the Xen-provided frame and skip the bare metal prologue.

Quoted 6 lines
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Fixes 1c3e5d3f60e2 ("x86/entry: Make entry_64_compat.S objtool clean")
#2 lore raw

[tip: x86/urgent] x86/entry: Assert that syscalls are on the right stack

In reply to Andy Lutomirski
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     c9c26150e61de441ab58b25c1f64afc049ee0fee
Gitweb:        https://git.kernel.org/tip/c9c26150e61de441ab58b25c1f64afc049ee0fee
Author:        Andy Lutomirski <[email protected]>
AuthorDate:    Fri, 26 Jun 2020 10:21:11 -07:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 01 Jul 2020 10:00:25 +02:00

x86/entry: Assert that syscalls are on the right stack

Now that the entry stack is a full page, it's too easy to regress the system call entry code and end up on the wrong stack without noticing. Assert that all system calls (SYSCALL64, SYSCALL32, SYSENTER, and INT80) are on the right stack and have pt_regs in the right place.

---
 arch/x86/entry/common.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)
arch/x86/entry/common.c +15 -3
index bd3f141..ed8ccc8 100644
@@ -45,6 +45,15 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/syscalls.h>
 
+/* Check that the stack and regs on entry from user mode are sane. */
+static void check_user_regs(struct pt_regs *regs)
+{
+	if (IS_ENABLED(CONFIG_DEBUG_ENTRY)) {
+		WARN_ON_ONCE(!on_thread_stack());
+		WARN_ON_ONCE(regs != task_pt_regs(current));
+	}
+}
+
 #ifdef CONFIG_CONTEXT_TRACKING
 /**
  * enter_from_user_mode - Establish state when coming from user mode
@@ -127,9 +136,6 @@ static long syscall_trace_enter(struct pt_regs *regs)
 	unsigned long ret = 0;
 	u32 work;
 
-	if (IS_ENABLED(CONFIG_DEBUG_ENTRY))
-		BUG_ON(regs != task_pt_regs(current));
-
 	work = READ_ONCE(ti->flags);
 
 	if (work & (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_EMU)) {
@@ -346,6 +352,8 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs)
 {
 	struct thread_info *ti;
 
+	check_user_regs(regs);
+
 	enter_from_user_mode();
 	instrumentation_begin();
 
@@ -409,6 +417,8 @@ static void do_syscall_32_irqs_on(struct pt_regs *regs)
 /* Handles int $0x80 */
 __visible noinstr void do_int80_syscall_32(struct pt_regs *regs)
 {
+	check_user_regs(regs);
+
 	enter_from_user_mode();
 	instrumentation_begin();
 
@@ -460,6 +470,8 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)
 					vdso_image_32.sym_int80_landing_pad;
 	bool success;
 
+	check_user_regs(regs);
+
 	/*
 	 * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
 	 * so that 'regs->ip -= 2' lands back on an int $0x80 instruction.
#5 lore raw

[tip: x86/urgent] x86/entry: Move SYSENTER's regs->sp and regs->flags fixups into C

In reply to Andy Lutomirski
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     d1721250f3ffed9afba3e1fb729947cec64c5a8a
Gitweb:        https://git.kernel.org/tip/d1721250f3ffed9afba3e1fb729947cec64c5a8a
Author:        Andy Lutomirski <[email protected]>
AuthorDate:    Fri, 26 Jun 2020 10:21:12 -07:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 01 Jul 2020 10:00:25 +02:00

x86/entry: Move SYSENTER's regs->sp and regs->flags fixups into C

The SYSENTER asm (32-bit and compat) contains fixups for regs->sp and regs->flags. Move the fixups into C and fix some comments while at it.

This is a valid cleanup all by itself, and it also simplifies the subsequent patch that will fix Xen PV SYSENTER.

---
 arch/x86/entry/common.c          | 12 ++++++++++++
 arch/x86/entry/entry_32.S        |  5 ++---
 arch/x86/entry/entry_64_compat.S | 11 +++++------
 3 files changed, 19 insertions(+), 9 deletions(-)
arch/x86/entry/common.c +12 -0
index ed8ccc8..f392a8b 100644
@@ -522,6 +522,18 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)
 		(regs->flags & (X86_EFLAGS_RF | X86_EFLAGS_TF | X86_EFLAGS_VM)) == 0;
 #endif
 }
+
+/* Returns 0 to return using IRET or 1 to return using SYSEXIT/SYSRETL. */
+__visible noinstr long do_SYSENTER_32(struct pt_regs *regs)
+{
+	/* SYSENTER loses RSP, but the vDSO saved it in RBP. */
+	regs->sp = regs->bp;
+
+	/* SYSENTER clobbers EFLAGS.IF.  Assume it was set in usermode. */
+	regs->flags |= X86_EFLAGS_IF;
+
+	return do_fast_syscall_32(regs);
+}
 #endif
 
 SYSCALL_DEFINE0(ni_syscall)
arch/x86/entry/entry_32.S +2 -3
index 024d7d2..2d0bd5d 100644
@@ -933,9 +933,8 @@ SYM_FUNC_START(entry_SYSENTER_32)
 
 .Lsysenter_past_esp:
 	pushl	$__USER_DS		/* pt_regs->ss */
-	pushl	%ebp			/* pt_regs->sp (stashed in bp) */
+	pushl	$0			/* pt_regs->sp (placeholder) */
 	pushfl				/* pt_regs->flags (except IF = 0) */
-	orl	$X86_EFLAGS_IF, (%esp)	/* Fix IF */
 	pushl	$__USER_CS		/* pt_regs->cs */
 	pushl	$0			/* pt_regs->ip = 0 (placeholder) */
 	pushl	%eax			/* pt_regs->orig_ax */
@@ -965,7 +964,7 @@ SYM_FUNC_START(entry_SYSENTER_32)
 .Lsysenter_flags_fixed:
 
 	movl	%esp, %eax
-	call	do_fast_syscall_32
+	call	do_SYSENTER_32
 	/* XEN PV guests always use IRET path */
 	ALTERNATIVE "testl %eax, %eax; jz .Lsyscall_32_done", \
 		    "jmp .Lsyscall_32_done", X86_FEATURE_XENPV
arch/x86/entry/entry_64_compat.S +5 -6
index 0f974ae..7b9d815 100644
@@ -68,16 +68,15 @@ SYM_CODE_START(entry_SYSENTER_compat)
 
 	/* Construct struct pt_regs on stack */
 	pushq	$__USER32_DS		/* pt_regs->ss */
-	pushq	%rbp			/* pt_regs->sp (stashed in bp) */
+	pushq	$0			/* pt_regs->sp = 0 (placeholder) */
 
 	/*
 	 * Push flags.  This is nasty.  First, interrupts are currently
-	 * off, but we need pt_regs->flags to have IF set.  Second, even
-	 * if TF was set when SYSENTER started, it's clear by now.  We fix
-	 * that later using TIF_SINGLESTEP.
+	 * off, but we need pt_regs->flags to have IF set.  Second, if TS
+	 * was set in usermode, it's still set, and we're singlestepping
+	 * through this code.  do_SYSENTER_32() will fix up IF.
 	 */
 	pushfq				/* pt_regs->flags (except IF = 0) */
-	orl	$X86_EFLAGS_IF, (%rsp)	/* Fix saved flags */
 	pushq	$__USER32_CS		/* pt_regs->cs */
 	pushq	$0			/* pt_regs->ip = 0 (placeholder) */
 	pushq	%rax			/* pt_regs->orig_ax */
@@ -135,7 +134,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
 .Lsysenter_flags_fixed:
 
 	movq	%rsp, %rdi
-	call	do_fast_syscall_32
+	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
#15 lore raw

[tip: x86/urgent] selftests/x86/syscall_nt: Clear weird flags after each test

In reply to Andy Lutomirski
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     a61fa2799ef9bf6c4f54cf7295036577cececc72
Gitweb:        https://git.kernel.org/tip/a61fa2799ef9bf6c4f54cf7295036577cececc72
Author:        Andy Lutomirski <[email protected]>
AuthorDate:    Fri, 26 Jun 2020 10:21:15 -07:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 01 Jul 2020 10:00:26 +02:00

selftests/x86/syscall_nt: Clear weird flags after each test

Clear the weird flags before logging to improve strace output -- logging results while, say, TF is set does no one any favors.

---
 tools/testing/selftests/x86/syscall_nt.c | 1 +
 1 file changed, 1 insertion(+)
tools/testing/selftests/x86/syscall_nt.c +1 -0
index f060534..5fc82b9 100644
@@ -59,6 +59,7 @@ static void do_it(unsigned long extraflags)
 	set_eflags(get_eflags() | extraflags);
 	syscall(SYS_getpid);
 	flags = get_eflags();
+	set_eflags(X86_EFLAGS_IF | X86_EFLAGS_FIXED);
 	if ((flags & extraflags) == extraflags) {
 		printf("[OK]\tThe syscall worked and flags are still set\n");
 	} else {
#8 lore raw

[tip: x86/urgent] x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

In reply to Andy Lutomirski
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     ffae641f57476369b4d503402b37ebe489d23395
Gitweb:        https://git.kernel.org/tip/ffae641f57476369b4d503402b37ebe489d23395
Author:        Andy Lutomirski <[email protected]>
AuthorDate:    Fri, 26 Jun 2020 10:21:13 -07:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 01 Jul 2020 10:00:26 +02:00

x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

The SYSENTER frame setup was nonsense. It worked by accident because the normal code into which the Xen asm jumped (entry_SYSENTER_32/compat) threw away SP without touching the stack. entry_SYSENTER_compat was recently modified such that it relied on having a valid stack pointer, so now the Xen asm needs to invoke it with a valid stack.

Fix it up like SYSCALL: use the Xen-provided frame and skip the bare metal prologue.

---
 arch/x86/entry/entry_64_compat.S |  1 +
 arch/x86/xen/xen-asm_64.S        | 20 ++++++++++++++++----
 2 files changed, 17 insertions(+), 4 deletions(-)
arch/x86/entry/entry_64_compat.S +1 -0
index 7b9d815..381a6de 100644
@@ -79,6 +79,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
 	pushfq				/* pt_regs->flags (except IF = 0) */
 	pushq	$__USER32_CS		/* pt_regs->cs */
 	pushq	$0			/* pt_regs->ip = 0 (placeholder) */
+SYM_INNER_LABEL(entry_SYSENTER_compat_after_hwframe, SYM_L_GLOBAL)
 	pushq	%rax			/* pt_regs->orig_ax */
 	pushq	%rdi			/* pt_regs->di */
 	pushq	%rsi			/* pt_regs->si */
arch/x86/xen/xen-asm_64.S +16 -4
index 5d252aa..e1e1c7e 100644
@@ -161,10 +161,22 @@ SYM_FUNC_END(xen_syscall32_target)
 
 /* 32-bit compat sysenter target */
 SYM_FUNC_START(xen_sysenter_target)
-	mov 0*8(%rsp), %rcx
-	mov 1*8(%rsp), %r11
-	mov 5*8(%rsp), %rsp
-	jmp entry_SYSENTER_compat
+	/*
+	 * NB: Xen is polite and clears TF from EFLAGS for us.  This means
+	 * that we don't need to guard against single step exceptions here.
+	 */
+	popq %rcx
+	popq %r11
+
+	/*
+	 * Neither Xen nor the kernel really knows what the old SS and
+	 * CS were.  The kernel expects __USER32_DS and __USER32_CS, so
+	 * report those values even though Xen will guess its own values.
+	 */
+	movq $__USER32_DS, 4*8(%rsp)
+	movq $__USER32_CS, 1*8(%rsp)
+
+	jmp entry_SYSENTER_compat_after_hwframe
 SYM_FUNC_END(xen_sysenter_target)
 
 #else /* !CONFIG_IA32_EMULATION */
#17 lore raw

[tip: x86/urgent] selftests/x86: Consolidate and fix get/set_eflags() helpers

In reply to Andy Lutomirski
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     cced0b24bb545bfe74fea96de84adc23c0146b05
Gitweb:        https://git.kernel.org/tip/cced0b24bb545bfe74fea96de84adc23c0146b05
Author:        Andy Lutomirski <[email protected]>
AuthorDate:    Fri, 26 Jun 2020 10:21:16 -07:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 01 Jul 2020 10:00:27 +02:00

selftests/x86: Consolidate and fix get/set_eflags() helpers

There are several copies of get_eflags() and set_eflags() and they all are buggy. Consolidate them and fix them. The fixes are:

Add memory clobbers. These are probably unnecessary but they make sure that the compiler doesn't move something past one of these calls when it shouldn't.

Respect the redzone on x86_64. There has no failure been observed related to this, but it's definitely a bug.

---
 tools/testing/selftests/x86/Makefile              |  4 +-
 tools/testing/selftests/x86/helpers.h             | 41 ++++++++++++++-
 tools/testing/selftests/x86/single_step_syscall.c | 17 +------
 tools/testing/selftests/x86/syscall_arg_fault.c   | 21 +-------
 tools/testing/selftests/x86/syscall_nt.c          | 20 +-------
 tools/testing/selftests/x86/test_vsyscall.c       | 15 +-----
 tools/testing/selftests/x86/unwind_vdso.c         | 23 +--------
 7 files changed, 51 insertions(+), 90 deletions(-)
 create mode 100644 tools/testing/selftests/x86/helpers.h
tools/testing/selftests/x86/Makefile +2 -2
index 5f16821..d2796ea 100644
@@ -70,10 +70,10 @@ all_64: $(BINARIES_64)
 
 EXTRA_CLEAN := $(BINARIES_32) $(BINARIES_64)
 
-$(BINARIES_32): $(OUTPUT)/%_32: %.c
+$(BINARIES_32): $(OUTPUT)/%_32: %.c helpers.h
 	$(CC) -m32 -o $@ $(CFLAGS) $(EXTRA_CFLAGS) $^ -lrt -ldl -lm
 
-$(BINARIES_64): $(OUTPUT)/%_64: %.c
+$(BINARIES_64): $(OUTPUT)/%_64: %.c helpers.h
 	$(CC) -m64 -o $@ $(CFLAGS) $(EXTRA_CFLAGS) $^ -lrt -ldl
 
 # x86_64 users should be encouraged to install 32-bit libraries
tools/testing/selftests/x86/helpers.h new file +41 -0

41 lines not on this page · raw message

tools/testing/selftests/x86/single_step_syscall.c +2 -15
index 1063328..120ac74 100644
@@ -31,6 +31,8 @@
 #include <sys/ptrace.h>
 #include <sys/user.h>
 
+#include "helpers.h"
+
 static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		       int flags)
 {
@@ -67,21 +69,6 @@ static unsigned char altstack_data[SIGSTKSZ];
 # define INT80_CLOBBERS
 #endif
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
-#define X86_EFLAGS_TF (1UL << 8)
-
 static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
 {
 	ucontext_t *ctx = (ucontext_t*)ctx_void;
tools/testing/selftests/x86/syscall_arg_fault.c +1 -20
index bc0ecc2..5b7abeb 100644
@@ -15,30 +15,11 @@
 #include <setjmp.h>
 #include <errno.h>
 
-#ifdef __x86_64__
-# define WIDTH "q"
-#else
-# define WIDTH "l"
-#endif
+#include "helpers.h"
 
 /* Our sigaltstack scratch space. */
 static unsigned char altstack_data[SIGSTKSZ];
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
-#define X86_EFLAGS_TF (1UL << 8)
-
 static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		       int flags)
 {
tools/testing/selftests/x86/syscall_nt.c +1 -19
index 5fc82b9..970e5e1 100644
@@ -13,29 +13,11 @@
 #include <signal.h>
 #include <err.h>
 #include <sys/syscall.h>
-#include <asm/processor-flags.h>
 
-#ifdef __x86_64__
-# define WIDTH "q"
-#else
-# define WIDTH "l"
-#endif
+#include "helpers.h"
 
 static unsigned int nerrs;
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
 static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		       int flags)
 {
tools/testing/selftests/x86/test_vsyscall.c +2 -13
index a4f4d4c..c41f24b 100644
@@ -20,6 +20,8 @@
 #include <setjmp.h>
 #include <sys/uio.h>
 
+#include "helpers.h"
+
 #ifdef __x86_64__
 # define VSYS(x) (x)
 #else
@@ -493,21 +495,8 @@ static int test_process_vm_readv(void)
 }
 
 #ifdef __x86_64__
-#define X86_EFLAGS_TF (1UL << 8)
 static volatile sig_atomic_t num_vsyscall_traps;
 
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushfq\n\tpopq %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("pushq %0\n\tpopfq" : : "rm" (eflags) : "flags");
-}
-
 static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
 {
 	ucontext_t *ctx = (ucontext_t *)ctx_void;
tools/testing/selftests/x86/unwind_vdso.c +2 -21
index 0075ccd..4c311e1 100644
@@ -11,6 +11,8 @@
 #include <features.h>
 #include <stdio.h>
 
+#include "helpers.h"
+
 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ < 16
 
 int main()
@@ -53,27 +55,6 @@ static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
 		err(1, "sigaction");
 }
 
-#ifdef __x86_64__
-# define WIDTH "q"
-#else
-# define WIDTH "l"
-#endif
-
-static unsigned long get_eflags(void)
-{
-	unsigned long eflags;
-	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
-	return eflags;
-}
-
-static void set_eflags(unsigned long eflags)
-{
-	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
-		      : : "rm" (eflags) : "flags");
-}
-
-#define X86_EFLAGS_TF (1UL << 8)
-
 static volatile sig_atomic_t nerrs;
 static unsigned long sysinfo;
 static bool got_sysinfo = false;
#13 lore raw

[tip: x86/urgent] selftests/x86/syscall_nt: Add more flag combinations

In reply to Andy Lutomirski
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     e4ef7de160c6b12639c4fc49bcacb25b860ac76d
Gitweb:        https://git.kernel.org/tip/e4ef7de160c6b12639c4fc49bcacb25b860ac76d
Author:        Andy Lutomirski <[email protected]>
AuthorDate:    Fri, 26 Jun 2020 10:21:14 -07:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 01 Jul 2020 10:00:26 +02:00

selftests/x86/syscall_nt: Add more flag combinations

Add EFLAGS.AC to the mix.

---
 tools/testing/selftests/x86/syscall_nt.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
tools/testing/selftests/x86/syscall_nt.c +15 -0
index 02309a1..f060534 100644
@@ -73,6 +73,12 @@ int main(void)
 	printf("[RUN]\tSet NT and issue a syscall\n");
 	do_it(X86_EFLAGS_NT);
 
+	printf("[RUN]\tSet AC and issue a syscall\n");
+	do_it(X86_EFLAGS_AC);
+
+	printf("[RUN]\tSet NT|AC and issue a syscall\n");
+	do_it(X86_EFLAGS_NT | X86_EFLAGS_AC);
+
 	/*
 	 * Now try it again with TF set -- TF forces returns via IRET in all
 	 * cases except non-ptregs-using 64-bit full fast path syscalls.
@@ -80,8 +86,17 @@ int main(void)
 
 	sethandler(SIGTRAP, sigtrap, 0);
 
+	printf("[RUN]\tSet TF and issue a syscall\n");
+	do_it(X86_EFLAGS_TF);
+
 	printf("[RUN]\tSet NT|TF and issue a syscall\n");
 	do_it(X86_EFLAGS_NT | X86_EFLAGS_TF);
 
+	printf("[RUN]\tSet AC|TF and issue a syscall\n");
+	do_it(X86_EFLAGS_AC | X86_EFLAGS_TF);
+
+	printf("[RUN]\tSet NT|AC|TF and issue a syscall\n");
+	do_it(X86_EFLAGS_NT | X86_EFLAGS_AC | X86_EFLAGS_TF);
+
 	return nerrs == 0 ? 0 : 1;
 }
#9 lore raw

Re: [PATCH 3/6] x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

In reply to Andy Lutomirski
Quoted 27 lines · Andy Lutomirski

The SYSENTER frame setup was nonsense. It worked by accident because the normal code into which the Xen asm jumped (entry_SYSENTER_32/compat) threw away SP without touching the stack. entry_SYSENTER_compat was recently modified such that it relied on having a valid stack pointer, so now the Xen asm needs to invoke it with a valid stack.

Fix it up like SYSCALL: use the Xen-provided frame and skip the bare metal prologue.

Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Fixes 1c3e5d3f60e2 ("x86/entry: Make entry_64_compat.S objtool clean")
---
 arch/x86/entry/entry_64_compat.S |  1 +
 arch/x86/xen/xen-asm_64.S        | 20 ++++++++++++++++----
 2 files changed, 17 insertions(+), 4 deletions(-)
arch/x86/entry/entry_64_compat.S +0 -0
index 7b9d8150f652..381a6de7de9c 100644

arch/x86/entry/entry_64_compat.S SYM_CODE_START(entry_SYSENTER_compat) · line 82

Quoted 4 lines
arch/x86/entry/entry_64_compat.S +1 -0
@@ -79,6 +79,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
        pushfq                          /* pt_regs->flags (except IF = 0) */
        pushq   $__USER32_CS            /* pt_regs->cs */
        pushq   $0                      /* pt_regs->ip = 0 (placeholder) */
+SYM_INNER_LABEL(entry_SYSENTER_compat_after_hwframe, SYM_L_GLOBAL)

This skips over the section that truncates the syscall number to 32-bits. The comments present some doubt that it is actually necessary, but the Xen path shouldn't differ from native. That code should be moved after this new label.

--
Brian Gerst

author maintainer
#10 lore raw

Re: [PATCH 3/6] x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

Quoted 24 lines · Brian Gerst
Quoted 22 lines · Andy Lutomirski

The SYSENTER frame setup was nonsense. It worked by accident because the normal code into which the Xen asm jumped (entry_SYSENTER_32/compat) threw away SP without touching the stack. entry_SYSENTER_compat was recently modified such that it relied on having a valid stack pointer, so now the Xen asm needs to invoke it with a valid stack.

Fix it up like SYSCALL: use the Xen-provided frame and skip the bare metal prologue.

Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Fixes 1c3e5d3f60e2 ("x86/entry: Make entry_64_compat.S objtool clean")
---
 arch/x86/entry/entry_64_compat.S |  1 +
 arch/x86/xen/xen-asm_64.S        | 20 ++++++++++++++++----
 2 files changed, 17 insertions(+), 4 deletions(-)
Quoted 14 lines

arch/x86/entry/entry_64_compat.S SYM_CODE_START(entry_SYSENTER_compat) · line 82

Quoted 9 lines
arch/x86/entry/entry_64_compat.S +1 -0
index 7b9d8150f652..381a6de7de9c 100644
@@ -79,6 +79,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
        pushfq                          /* pt_regs->flags (except IF = 0) */
        pushq   $__USER32_CS            /* pt_regs->cs */
        pushq   $0                      /* pt_regs->ip = 0 (placeholder) */
+SYM_INNER_LABEL(entry_SYSENTER_compat_after_hwframe, SYM_L_GLOBAL)

This skips over the section that truncates the syscall number to 32-bits. The comments present some doubt that it is actually necessary, but the Xen path shouldn't differ from native. That code should be moved after this new label.

Whoops. I thought I caught that myself, but apparently not. I'll fix it.

Quoted 3 lines

--
Brian Gerst

#11 lore raw

Re: [PATCH 3/6] x86/entry/64/compat: Fix Xen PV SYSENTER frame setup

Quoted 39 lines · Andy Lutomirski
Quoted 23 lines · Brian Gerst
Quoted 22 lines · Andy Lutomirski

The SYSENTER frame setup was nonsense. It worked by accident because the normal code into which the Xen asm jumped (entry_SYSENTER_32/compat) threw away SP without touching the stack. entry_SYSENTER_compat was recently modified such that it relied on having a valid stack pointer, so now the Xen asm needs to invoke it with a valid stack.

Fix it up like SYSCALL: use the Xen-provided frame and skip the bare metal prologue.

Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Fixes 1c3e5d3f60e2 ("x86/entry: Make entry_64_compat.S objtool clean")
---
 arch/x86/entry/entry_64_compat.S |  1 +
 arch/x86/xen/xen-asm_64.S        | 20 ++++++++++++++++----
 2 files changed, 17 insertions(+), 4 deletions(-)
Quoted 14 lines

arch/x86/entry/entry_64_compat.S SYM_CODE_START(entry_SYSENTER_compat) · line 82

Quoted 9 lines
arch/x86/entry/entry_64_compat.S +1 -0
index 7b9d8150f652..381a6de7de9c 100644
@@ -79,6 +79,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
        pushfq                          /* pt_regs->flags (except IF = 0) */
        pushq   $__USER32_CS            /* pt_regs->cs */
        pushq   $0                      /* pt_regs->ip = 0 (placeholder) */
+SYM_INNER_LABEL(entry_SYSENTER_compat_after_hwframe, SYM_L_GLOBAL)

This skips over the section that truncates the syscall number to 32-bits. The comments present some doubt that it is actually necessary, but the Xen path shouldn't differ from native. That code should be moved after this new label.

Quoted 1 line

Whoops. I thought I caught that myself, but apparently not. I'll fix it.

Darn. I already applied that lot. Can you please send a delta fix?

Thanks,

tglx

#3 lore raw

PATCH 1/1 x86/entry: Mark check_user_regs() noinstr

In reply to Andy Lutomirski
Quoted 1 line · Andy Lutomirski

+static void check_user_regs(struct pt_regs *regs)

---
Subject: x86/entry: Mark check_user_regs() noinstr

vmlinux.o: warning: objtool: do_syscall_64()+0xb: call to check_user_regs() leaves .noinstr.text section vmlinux.o: warning: objtool: do_int80_syscall_32()+0x4: call to check_user_regs() leaves .noinstr.text section vmlinux.o: warning: objtool: do_fast_syscall_32()+0x25: call to check_user_regs() leaves .noinstr.text section vmlinux.o: warning: objtool: idtentry_enter_cond_rcu()+0x3d: call to check_user_regs() leaves .noinstr.text section vmlinux.o: warning: objtool: idtentry_enter_user()+0x0: call to check_user_regs() leaves .noinstr.text section

---
 arch/x86/entry/common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/entry/common.c +1 -1
index e83b3f14897c..ea7b515e3bc2 100644
@@ -46,7 +46,7 @@
 #include <trace/events/syscalls.h>
 
 /* Check that the stack and regs on entry from user mode are sane. */
-static void check_user_regs(struct pt_regs *regs)
+static noinstr void check_user_regs(struct pt_regs *regs)
 {
 	if (IS_ENABLED(CONFIG_DEBUG_ENTRY)) {
 		/*