entry, x86, kvm: Generic entry/exit functionality for host and guest
Subsystems (12)
v5 cover entry, x86, kvm: Generic entry/exit functionality for host and guest
This is the 5th version of generic entry/exit functionality for host and guest.
The 4th version is available here:
https://lore.kernel.org/r/[email protected]
Changes vs. V4:
- Add the missing instrumentation prevetions to the entry Makefile (Kees)
- Rename exit_to_guest to xfer_to_guest (Sean)
The patches depend on:
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86/entry
The lot is also available from git:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git x86/entry
Thanks,
tglx
PATCH v5 1/15 seccomp: Provide stub for __secure_computing()
From: Thomas Gleixner <[email protected]>
To avoid #ifdeffery in the upcoming generic syscall entry work code provide a stub for __secure_computing() as this is preferred over secure_computing() because the TIF flag is already evaluated.
Signed-off-by Thomas Gleixner <[email protected]>
Acked-by Kees Cook <[email protected]>
---
V4: New patch
--- include/linux/seccomp.h | 1 + 1 file changed, 1 insertion(+)
include/linux/seccomp.h +1 -0
@@ -61,6 +61,7 @@ struct seccomp_filter { };#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER static inline int secure_computing (void ) {return 0 ; } +static inline int __secure_computing (void ) {return 0 ; }#else static inline void secure_computing_strict (int this_syscall) {return ; }#endif
[tip: core/entry] seccomp: Provide stub for __secure_computing()
The following commit has been merged into the core/entry branch of tip:
The following commit has been merged into the core/entry branch of tip:
Commit-ID: 6823ecabf03031d610a6c5afe7ed4b4fd659a99f Gitweb: https://git.kernel.org/tip/6823ecabf03031d610a6c5afe7ed4b4fd659a99f Author: Thomas Gleixner <[email protected]> AuthorDate: Wed, 22 Jul 2020 23:59:55 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 14:59:03 +02:00
seccomp: Provide stub for __secure_computing()
To avoid #ifdeffery in the upcoming generic syscall entry work code provide a stub for __secure_computing() as this is preferred over secure_computing() because the TIF flag is already evaluated.
Signed-off-by Thomas Gleixner <[email protected]>
Acked-by Kees Cook <[email protected]>
--- include/linux/seccomp.h | 1 + 1 file changed, 1 insertion(+)
include/linux/seccomp.h +1 -0
@@ -61,6 +61,7 @@ struct seccomp_filter { };#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER static inline int secure_computing (void ) {return 0 ; } +static inline int __secure_computing (void ) {return 0 ; }#else static inline void secure_computing_strict (int this_syscall) {return ; }#endif
PATCH v5 2/15 entry: Provide generic syscall entry functionality
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
On syscall entry certain work needs to be done:
- Establish state (lockdep, context tracking, tracing)
- Conditional work (ptrace, seccomp, audit...)
This code is needlessly duplicated and different in all architectures.
Provide a generic version based on the x86 implementation which has all the RCU and instrumentation bits right.
As interrupt/exception entry from user space needs parts of the same functionality, provide a function for this as well.
syscall_enter_from_user_mode() and irqentry_enter_from_user_mode() must be called right after the low level ASM entry. The calling code must be non-instrumentable. After the functions returns state is correct and the subsequent functions can be instrumented.
Signed-off-by Thomas Gleixner <[email protected]>
Acked-by Kees Cook <[email protected]>
---
V5: Add the missing protections against sanitizers and other goodies
V4: Remove the architecture wrappers for seccomp and audit and use
the generic code (Kees)
Fix TIF_SYSCALL_AUDIT dummy define and remove TIF_SYSCALL_TRACE as it's defined in all architectures.
V3: Adapt to noinstr changes. Add interrupt/exception entry function.
V2: Fix function documentation (Mike)
Add comment about return value (Andy)
--- arch/Kconfig | 3 + include/linux/entry-common.h | 121 +++++++++++++++++++++++++++++++++++++++++++ kernel/Makefile | 1 kernel/entry/Makefile | 12 ++++ kernel/entry/common.c | 88 +++++++++++++++++++++++++++++++ 5 files changed, 225 insertions(+)
arch/Kconfig +3 -0
@@ -27,6 +27,9 @@ config HAVE_IMA_KEXEC config HOTPLUG_SMT bool +config GENERIC_ENTRY + bool + config OPROFILE tristate "OProfile system profiling" depends on PROFILING
include/linux/entry-common.h new file +121 -0
121 lines not on this page · raw message
kernel/Makefile +1 -0
@@ -48,6 +48,7 @@ obj-y += irq/ obj-y += rcu/ obj-y += livepatch/ obj-y += dma/ +obj-y += entry/ obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o obj-$(CONFIG_FREEZER) += freezer.o
kernel/entry/Makefile new file +12 -0
@@ -0,0 +1,12 @@ +# SPDX-License-Identifier: GPL-2.0 + +# Prevent the noinstr section from being pestered by sanitizer and other goodies +# as long as these things cannot be disabled per function. +KASAN_SANITIZE := n +UBSAN_SANITIZE := n +KCOV_INSTRUMENT := n + +CFLAGS_REMOVE_common.o = -fstack-protector -fstack-protector-strong +CFLAGS_common.o += -fno-stack-protector + +obj-$(CONFIG_GENERIC_ENTRY) += common.o
kernel/entry/common.c new file +88 -0
88 lines not on this page · raw message
[tip: core/entry] entry: Provide generic syscall entry functionality
The following commit has been merged into the core/entry branch of tip:
The following commit has been merged into the core/entry branch of tip:
Commit-ID: 142781e108b13b2b0e8f035cfb5bfbbc8f14d887 Gitweb: https://git.kernel.org/tip/142781e108b13b2b0e8f035cfb5bfbbc8f14d887 Author: Thomas Gleixner <[email protected]> AuthorDate: Wed, 22 Jul 2020 23:59:56 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 14:59:03 +02:00
entry: Provide generic syscall entry functionality
On syscall entry certain work needs to be done:
- Establish state (lockdep, context tracking, tracing)
- Conditional work (ptrace, seccomp, audit...)
This code is needlessly duplicated and different in all architectures.
Provide a generic version based on the x86 implementation which has all the RCU and instrumentation bits right.
As interrupt/exception entry from user space needs parts of the same functionality, provide a function for this as well.
syscall_enter_from_user_mode() and irqentry_enter_from_user_mode() must be called right after the low level ASM entry. The calling code must be non-instrumentable. After the functions returns state is correct and the subsequent functions can be instrumented.
Signed-off-by Thomas Gleixner <[email protected]>
Acked-by Kees Cook <[email protected]>
--- arch/Kconfig | 3 +- include/linux/entry-common.h | 121 ++++++++++++++++++++++++++++++++++- kernel/Makefile | 1 +- kernel/entry/Makefile | 12 +++- kernel/entry/common.c | 88 +++++++++++++++++++++++++- 5 files changed, 225 insertions(+) create mode 100644 include/linux/entry-common.h create mode 100644 kernel/entry/Makefile create mode 100644 kernel/entry/common.c
arch/Kconfig +3 -0
@@ -27,6 +27,9 @@ config HAVE_IMA_KEXEC config HOTPLUG_SMT bool +config GENERIC_ENTRY + bool + config OPROFILE tristate "OProfile system profiling" depends on PROFILING
include/linux/entry-common.h new file +121 -0
121 lines not on this page · raw message
kernel/Makefile +1 -0
@@ -48,6 +48,7 @@ obj-y += irq/ obj-y += rcu/ obj-y += livepatch/ obj-y += dma/ +obj-y += entry/ obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o obj-$(CONFIG_FREEZER) += freezer.o
kernel/entry/Makefile new file +12 -0
@@ -0,0 +1,12 @@ +# SPDX-License-Identifier: GPL-2.0 + +# Prevent the noinstr section from being pestered by sanitizer and other goodies +# as long as these things cannot be disabled per function. +KASAN_SANITIZE := n +UBSAN_SANITIZE := n +KCOV_INSTRUMENT := n + +CFLAGS_REMOVE_common.o = -fstack-protector -fstack-protector-strong +CFLAGS_common.o += -fno-stack-protector + +obj-$(CONFIG_GENERIC_ENTRY) += common.o
kernel/entry/common.c new file +88 -0
88 lines not on this page · raw message
PATCH v5 3/15 entry: Provide generic syscall exit function
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Like syscall entry all architectures have similar and pointlessly different code to handle pending work before returning from a syscall to user space.
1) One-time syscall exit work:
- rseq syscall exit
- audit
- syscall tracing
- tracehook (single stepping)
2) Preparatory work
- Exit to user mode loop (common TIF handling).
- Architecture specific one time work arch_exit_to_user_mode_prepare()
- Address limit and lockdep checks
3) Final transition (lockdep, tracing, context tracking, RCU). Invokes arch_exit_to_user_mode() to handle e.g. speculation mitigations
Provide a generic version based on the x86 code which has all the RCU and instrumentation protections right.
Provide a variant for interrupt return to user mode as well which shares the above #2 and #3 work items.
After syscall_exit_to_user_mode() and irqentry_exit_to_user_mode() the architecture code just has to return to user space. The code after returning from these functions must not be instrumented.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
---
V4: Remove _TIF_NOTIFY_RESUME dummy define as it is defined by all architectures
Remove the return value helper and use the existing regs_return_value()
--- include/linux/entry-common.h | 189 +++++++++++++++++++++++++++++++++++++++++++ kernel/entry/common.c | 169 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 358 insertions(+)
include/linux/entry-common.h +189 -0
@@ -29,6 +29,14 @@# define _TIF_SYSCALL_AUDIT (0)#endif +#ifndef _TIF_PATCH_PENDING +# define _TIF_PATCH_PENDING (0) +#endif + +#ifndef _TIF_UPROBE +# define _TIF_UPROBE (0) +#endif +/* * TIF flags handled in syscall_enter_from_usermode() */
@@ -41,6 +49,29 @@ _TIF_SYSCALL_TRACEPOINT | _TIF_SYSCALL_EMU | \ ARCH_SYSCALL_ENTER_WORK) +/* +* TIF flags handled in syscall_exit_to_user_mode() +*/ +#ifndef ARCH_SYSCALL_EXIT_WORK +# define ARCH_SYSCALL_EXIT_WORK (0) +#endif + +#define SYSCALL_EXIT_WORK \ + (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | \ + _TIF_SYSCALL_TRACEPOINT | ARCH_SYSCALL_EXIT_WORK) + +/* +* TIF flags handled in exit_to_user_mode_loop() +*/ +#ifndef ARCH_EXIT_TO_USER_MODE_WORK +# define ARCH_EXIT_TO_USER_MODE_WORK (0) +#endif + +#define EXIT_TO_USER_MODE_WORK \ + (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \ + _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | \ + ARCH_EXIT_TO_USER_MODE_WORK) +/** * arch_check_user_regs - Architecture specific sanity check for user mode regs * @regs: Pointer to currents pt_regs
@@ -106,6 +137,149 @@ static inline __must_check int arch_sysclong syscall_enter_from_user_mode (struct pt_regs *regs,long syscall);/** +* local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable() +* @ti_work: Cached TIF flags gathered with interrupts disabled +* +* Defaults to local_irq_enable(). Can be supplied by architecture specific +* code. +*/ +static inline void local_irq_enable_exit_to_user (unsigned long ti_work); + +#ifndef local_irq_enable_exit_to_user +static inline void local_irq_enable_exit_to_user (unsigned long ti_work) +{ +local_irq_enable (); +} +#endif + +/** +* local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable() +* +* Defaults to local_irq_disable(). Can be supplied by architecture specific +* code. +*/ +static inline void local_irq_disable_exit_to_user (void ); + +#ifndef local_irq_disable_exit_to_user +static inline void local_irq_disable_exit_to_user (void ) +{ +local_irq_disable (); +} +#endif + +/** +* arch_exit_to_user_mode_work - Architecture specific TIF work for exit +* to user mode. +* @regs: Pointer to currents pt_regs +* @ti_work: Cached TIF flags gathered with interrupts disabled +* +* Invoked from exit_to_user_mode_loop() with interrupt enabled +* +* Defaults to NOOP. Can be supplied by architecture specific code. +*/ +static inline void arch_exit_to_user_mode_work (struct pt_regs *regs, +unsigned long ti_work); + +#ifndef arch_exit_to_user_mode_work +static inline void arch_exit_to_user_mode_work (struct pt_regs *regs, +unsigned long ti_work) +{ +} +#endif + +/** +* arch_exit_to_user_mode_prepare - Architecture specific preparation for +* exit to user mode. +* @regs: Pointer to currents pt_regs +* @ti_work: Cached TIF flags gathered with interrupts disabled +* +* Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last +* function before return. Defaults to NOOP. +*/ +static inline void arch_exit_to_user_mode_prepare (struct pt_regs *regs, +unsigned long ti_work); + +#ifndef arch_exit_to_user_mode_prepare +static inline void arch_exit_to_user_mode_prepare (struct pt_regs *regs, +unsigned long ti_work) +{ +} +#endif + +/** +* arch_exit_to_user_mode - Architecture specific final work before +* exit to user mode. +* +* Invoked from exit_to_user_mode() with interrupt disabled as the last +* function before return. Defaults to NOOP. +* +* This needs to be __always_inline because it is non-instrumentable code +* invoked after context tracking switched to user mode. +* +* An architecture implementation must not do anything complex, no locking +* etc. The main purpose is for speculation mitigations. +*/ +static __always_inline void arch_exit_to_user_mode(void); + +#ifndef arch_exit_to_user_mode +static __always_inline void arch_exit_to_user_mode(void) { } +#endif + +/** +* arch_do_signal - Architecture specific signal delivery function +* @regs: Pointer to currents pt_regs +* +* Invoked from exit_to_user_mode_loop(). +*/ +void arch_do_signal (struct pt_regs *regs); + +/** +* arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit() +* @regs: Pointer to currents pt_regs +* @step: Indicator for single step +* +* Defaults to tracehook_report_syscall_exit(). Can be replaced by +* architecture specific code. +* +* Invoked from syscall_exit_to_user_mode() +*/ +static inline void arch_syscall_exit_tracehook (struct pt_regs *regs,bool step); + +#ifndef arch_syscall_exit_tracehook +static inline void arch_syscall_exit_tracehook (struct pt_regs *regs,bool step) +{ +tracehook_report_syscall_exit (regs, step); +} +#endif + +/** +* syscall_exit_to_user_mode - Handle work before returning to user mode +* @regs: Pointer to currents pt_regs +* +* Invoked with interrupts enabled and fully valid regs. Returns with all +* work handled, interrupts disabled such that the caller can immediately +* switch to user mode. Called from architecture specific syscall and ret +* from fork code. +* +* The call order is: +* 1) One-time syscall exit work: +* - rseq syscall exit +* - audit +* - syscall tracing +* - tracehook (single stepping) +* +* 2) Preparatory work +* - Exit to user mode loop (common TIF handling). Invokes +* arch_exit_to_user_mode_work() for architecture specific TIF work +* - Architecture specific one time work arch_exit_to_user_mode_prepare() +* - Address limit and lockdep checks +* +* 3) Final transition (lockdep, tracing, context tracking, RCU). Invokes +* arch_exit_to_user_mode() to handle e.g. speculation mitigations +*/ +void syscall_exit_to_user_mode (struct pt_regs *regs); + +/** * irqentry_enter_from_user_mode - Establish state before invoking the irq handler * @regs: Pointer to currents pt_regs *
@@ -118,4 +292,19 @@ long syscall_enter_from_user_mode(struct*/ void irqentry_enter_from_user_mode (struct pt_regs *regs); +/** +* irqentry_exit_to_user_mode - Interrupt exit work +* @regs: Pointer to current's pt_regs +* +* Invoked with interrupts disbled and fully valid regs. Returns with all +* work handled, interrupts disabled such that the caller can immediately +* switch to user mode. Called from architecture specific interrupt +* handling code. +* +* The call order is #2 and #3 as described in syscall_exit_to_user_mode(). +* Interrupt exit is not invoking #1 which is the syscall specific one time +* work. +*/ +void irqentry_exit_to_user_mode (struct pt_regs *regs); +#endif
kernel/entry/common.c +169 -0
182 lines not on this page · raw message
[tip: core/entry] entry: Provide generic syscall exit function
The following commit has been merged into the core/entry branch of tip:
The following commit has been merged into the core/entry branch of tip:
Commit-ID: a9f3a74a29af095f3e1b89e9176f8127912ae0f0 Gitweb: https://git.kernel.org/tip/a9f3a74a29af095f3e1b89e9176f8127912ae0f0 Author: Thomas Gleixner <[email protected]> AuthorDate: Wed, 22 Jul 2020 23:59:57 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 14:59:03 +02:00
entry: Provide generic syscall exit function
Like syscall entry all architectures have similar and pointlessly different code to handle pending work before returning from a syscall to user space.
1) One-time syscall exit work:
- rseq syscall exit
- audit
- syscall tracing
- tracehook (single stepping)
2) Preparatory work
- Exit to user mode loop (common TIF handling).
- Architecture specific one time work arch_exit_to_user_mode_prepare()
- Address limit and lockdep checks
3) Final transition (lockdep, tracing, context tracking, RCU). Invokes arch_exit_to_user_mode() to handle e.g. speculation mitigations
Provide a generic version based on the x86 code which has all the RCU and instrumentation protections right.
Provide a variant for interrupt return to user mode as well which shares the above #2 and #3 work items.
After syscall_exit_to_user_mode() and irqentry_exit_to_user_mode() the architecture code just has to return to user space. The code after returning from these functions must not be instrumented.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- include/linux/entry-common.h | 189 ++++++++++++++++++++++++++++++++++- kernel/entry/common.c | 169 ++++++++++++++++++++++++++++++- 2 files changed, 358 insertions(+)
include/linux/entry-common.h +189 -0
@@ -29,6 +29,14 @@# define _TIF_SYSCALL_AUDIT (0)#endif +#ifndef _TIF_PATCH_PENDING +# define _TIF_PATCH_PENDING (0) +#endif + +#ifndef _TIF_UPROBE +# define _TIF_UPROBE (0) +#endif +/* * TIF flags handled in syscall_enter_from_usermode() */
@@ -41,6 +49,29 @@ _TIF_SYSCALL_TRACEPOINT | _TIF_SYSCALL_EMU | \ ARCH_SYSCALL_ENTER_WORK) +/* +* TIF flags handled in syscall_exit_to_user_mode() +*/ +#ifndef ARCH_SYSCALL_EXIT_WORK +# define ARCH_SYSCALL_EXIT_WORK (0) +#endif + +#define SYSCALL_EXIT_WORK \ + (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | \ + _TIF_SYSCALL_TRACEPOINT | ARCH_SYSCALL_EXIT_WORK) + +/* +* TIF flags handled in exit_to_user_mode_loop() +*/ +#ifndef ARCH_EXIT_TO_USER_MODE_WORK +# define ARCH_EXIT_TO_USER_MODE_WORK (0) +#endif + +#define EXIT_TO_USER_MODE_WORK \ + (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \ + _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | \ + ARCH_EXIT_TO_USER_MODE_WORK) +/** * arch_check_user_regs - Architecture specific sanity check for user mode regs * @regs: Pointer to currents pt_regs
@@ -106,6 +137,149 @@ static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regslong syscall_enter_from_user_mode (struct pt_regs *regs,long syscall);/** +* local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable() +* @ti_work: Cached TIF flags gathered with interrupts disabled +* +* Defaults to local_irq_enable(). Can be supplied by architecture specific +* code. +*/ +static inline void local_irq_enable_exit_to_user (unsigned long ti_work); + +#ifndef local_irq_enable_exit_to_user +static inline void local_irq_enable_exit_to_user (unsigned long ti_work) +{ +local_irq_enable (); +} +#endif + +/** +* local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable() +* +* Defaults to local_irq_disable(). Can be supplied by architecture specific +* code. +*/ +static inline void local_irq_disable_exit_to_user (void ); + +#ifndef local_irq_disable_exit_to_user +static inline void local_irq_disable_exit_to_user (void ) +{ +local_irq_disable (); +} +#endif + +/** +* arch_exit_to_user_mode_work - Architecture specific TIF work for exit +* to user mode. +* @regs: Pointer to currents pt_regs +* @ti_work: Cached TIF flags gathered with interrupts disabled +* +* Invoked from exit_to_user_mode_loop() with interrupt enabled +* +* Defaults to NOOP. Can be supplied by architecture specific code. +*/ +static inline void arch_exit_to_user_mode_work (struct pt_regs *regs, +unsigned long ti_work); + +#ifndef arch_exit_to_user_mode_work +static inline void arch_exit_to_user_mode_work (struct pt_regs *regs, +unsigned long ti_work) +{ +} +#endif + +/** +* arch_exit_to_user_mode_prepare - Architecture specific preparation for +* exit to user mode. +* @regs: Pointer to currents pt_regs +* @ti_work: Cached TIF flags gathered with interrupts disabled +* +* Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last +* function before return. Defaults to NOOP. +*/ +static inline void arch_exit_to_user_mode_prepare (struct pt_regs *regs, +unsigned long ti_work); + +#ifndef arch_exit_to_user_mode_prepare +static inline void arch_exit_to_user_mode_prepare (struct pt_regs *regs, +unsigned long ti_work) +{ +} +#endif + +/** +* arch_exit_to_user_mode - Architecture specific final work before +* exit to user mode. +* +* Invoked from exit_to_user_mode() with interrupt disabled as the last +* function before return. Defaults to NOOP. +* +* This needs to be __always_inline because it is non-instrumentable code +* invoked after context tracking switched to user mode. +* +* An architecture implementation must not do anything complex, no locking +* etc. The main purpose is for speculation mitigations. +*/ +static __always_inline void arch_exit_to_user_mode(void); + +#ifndef arch_exit_to_user_mode +static __always_inline void arch_exit_to_user_mode(void) { } +#endif + +/** +* arch_do_signal - Architecture specific signal delivery function +* @regs: Pointer to currents pt_regs +* +* Invoked from exit_to_user_mode_loop(). +*/ +void arch_do_signal (struct pt_regs *regs); + +/** +* arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit() +* @regs: Pointer to currents pt_regs +* @step: Indicator for single step +* +* Defaults to tracehook_report_syscall_exit(). Can be replaced by +* architecture specific code. +* +* Invoked from syscall_exit_to_user_mode() +*/ +static inline void arch_syscall_exit_tracehook (struct pt_regs *regs,bool step); + +#ifndef arch_syscall_exit_tracehook +static inline void arch_syscall_exit_tracehook (struct pt_regs *regs,bool step) +{ +tracehook_report_syscall_exit (regs, step); +} +#endif + +/** +* syscall_exit_to_user_mode - Handle work before returning to user mode +* @regs: Pointer to currents pt_regs +* +* Invoked with interrupts enabled and fully valid regs. Returns with all +* work handled, interrupts disabled such that the caller can immediately +* switch to user mode. Called from architecture specific syscall and ret +* from fork code. +* +* The call order is: +* 1) One-time syscall exit work: +* - rseq syscall exit +* - audit +* - syscall tracing +* - tracehook (single stepping) +* +* 2) Preparatory work +* - Exit to user mode loop (common TIF handling). Invokes +* arch_exit_to_user_mode_work() for architecture specific TIF work +* - Architecture specific one time work arch_exit_to_user_mode_prepare() +* - Address limit and lockdep checks +* +* 3) Final transition (lockdep, tracing, context tracking, RCU). Invokes +* arch_exit_to_user_mode() to handle e.g. speculation mitigations +*/ +void syscall_exit_to_user_mode (struct pt_regs *regs); + +/** * irqentry_enter_from_user_mode - Establish state before invoking the irq handler * @regs: Pointer to currents pt_regs *
@@ -118,4 +292,19 @@ long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall);*/ void irqentry_enter_from_user_mode (struct pt_regs *regs); +/** +* irqentry_exit_to_user_mode - Interrupt exit work +* @regs: Pointer to current's pt_regs +* +* Invoked with interrupts disbled and fully valid regs. Returns with all +* work handled, interrupts disabled such that the caller can immediately +* switch to user mode. Called from architecture specific interrupt +* handling code. +* +* The call order is #2 and #3 as described in syscall_exit_to_user_mode(). +* Interrupt exit is not invoking #1 which is the syscall specific one time +* work. +*/ +void irqentry_exit_to_user_mode (struct pt_regs *regs); +#endif
kernel/entry/common.c +169 -0
182 lines not on this page · raw message
PATCH v5 4/15 entry: Provide generic interrupt entry/exit code
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Like the syscall entry/exit code interrupt/exception entry after the real low level ASM bits should not be different accross architectures.
Provide a generic version based on the x86 code.
irqentry_enter() is called after the low level entry code and irqentry_exit() must be invoked right before returning to the low level code which just contains the actual return logic. The code before irqentry_enter() and irqentry_exit() must not be instrumented. Code after irqentry_enter() and before irqentry_exit() can be instrumented.
irqentry_enter() invokes irqentry_enter_from_user_mode() if the interrupt/exception came from user mode. If if entered from kernel mode it handles the kernel mode variant of establishing state for lockdep, RCU and tracing depending on the kernel context it interrupted (idle, non-idle).
Signed-off-by Thomas Gleixner <[email protected]>
--- include/linux/entry-common.h | 62 ++++++++++++++++++++++ kernel/entry/common.c | 117 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 179 insertions(+)
include/linux/entry-common.h +62 -0
@@ -307,4 +307,66 @@ void irqentry_enter_from_user_mode(struc*/ void irqentry_exit_to_user_mode (struct pt_regs *regs); +#ifndef irqentry_state +typedef struct irqentry_state { +bool exit_rcu; +}irqentry_state_t ; +#endif + +/** +* irqentry_enter - Handle state tracking on ordinary interrupt entries +* @regs: Pointer to pt_regs of interrupted context +* +* Invokes: +* - lockdep irqflag state tracking as low level ASM entry disabled +* interrupts. +* +* - Context tracking if the exception hit user mode. +* +* - The hardirq tracer to keep the state consistent as low level ASM +* entry disabled interrupts. +* +* As a precondition, this requires that the entry came from user mode, +* idle, or a kernel context in which RCU is watching. +* +* For kernel mode entries RCU handling is done conditional. If RCU is +* watching then the only RCU requirement is to check whether the tick has +* to be restarted. If RCU is not watching then rcu_irq_enter() has to be +* invoked on entry and rcu_irq_exit() on exit. +* +* Avoiding the rcu_irq_enter/exit() calls is an optimization but also +* solves the problem of kernel mode pagefaults which can schedule, which +* is not possible after invoking rcu_irq_enter() without undoing it. +* +* For user mode entries irqentry_enter_from_user_mode() is invoked to +* establish the proper context for NOHZ_FULL. Otherwise scheduling on exit +* would not be possible. +* +* Returns: An opaque object that must be passed to idtentry_exit() +*/ +irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs); + +/** +* irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt +* +* Conditional reschedule with additional sanity checks. +*/ +void irqentry_exit_cond_resched (void ); + +/** +* irqentry_exit - Handle return from exception that used irqentry_enter() +* @regs: Pointer to pt_regs (exception entry regs) +* @state: Return value from matching call to irqentry_enter() +* +* Depending on the return target (kernel/user) this runs the necessary +* preemption and work checks if possible and reguired and returns to +* the caller with interrupts disabled and no further work pending. +* +* This is the last action before returning to the low level ASM code which +* just needs to return to the appropriate context. +* +* Counterpart to irqentry_enter(). +*/ +void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state); +#endif
kernel/entry/common.c +117 -0
120 lines not on this page · raw message
[tip: core/entry] entry: Provide generic interrupt entry/exit code
The following commit has been merged into the core/entry branch of tip:
The following commit has been merged into the core/entry branch of tip:
Commit-ID: a5497bab5f72dce38a259a53fd3ac1239a7ecf40 Gitweb: https://git.kernel.org/tip/a5497bab5f72dce38a259a53fd3ac1239a7ecf40 Author: Thomas Gleixner <[email protected]> AuthorDate: Wed, 22 Jul 2020 23:59:58 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 14:59:04 +02:00
entry: Provide generic interrupt entry/exit code
Like the syscall entry/exit code interrupt/exception entry after the real low level ASM bits should not be different accross architectures.
Provide a generic version based on the x86 code.
irqentry_enter() is called after the low level entry code and irqentry_exit() must be invoked right before returning to the low level code which just contains the actual return logic. The code before irqentry_enter() and irqentry_exit() must not be instrumented. Code after irqentry_enter() and before irqentry_exit() can be instrumented.
irqentry_enter() invokes irqentry_enter_from_user_mode() if the interrupt/exception came from user mode. If if entered from kernel mode it handles the kernel mode variant of establishing state for lockdep, RCU and tracing depending on the kernel context it interrupted (idle, non-idle).
Signed-off-by Thomas Gleixner <[email protected]>
--- include/linux/entry-common.h | 62 ++++++++++++++++++- kernel/entry/common.c | 117 ++++++++++++++++++++++++++++++++++- 2 files changed, 179 insertions(+)
include/linux/entry-common.h +62 -0
@@ -307,4 +307,66 @@ void irqentry_enter_from_user_mode(struct pt_regs *regs);*/ void irqentry_exit_to_user_mode (struct pt_regs *regs); +#ifndef irqentry_state +typedef struct irqentry_state { +bool exit_rcu; +}irqentry_state_t ; +#endif + +/** +* irqentry_enter - Handle state tracking on ordinary interrupt entries +* @regs: Pointer to pt_regs of interrupted context +* +* Invokes: +* - lockdep irqflag state tracking as low level ASM entry disabled +* interrupts. +* +* - Context tracking if the exception hit user mode. +* +* - The hardirq tracer to keep the state consistent as low level ASM +* entry disabled interrupts. +* +* As a precondition, this requires that the entry came from user mode, +* idle, or a kernel context in which RCU is watching. +* +* For kernel mode entries RCU handling is done conditional. If RCU is +* watching then the only RCU requirement is to check whether the tick has +* to be restarted. If RCU is not watching then rcu_irq_enter() has to be +* invoked on entry and rcu_irq_exit() on exit. +* +* Avoiding the rcu_irq_enter/exit() calls is an optimization but also +* solves the problem of kernel mode pagefaults which can schedule, which +* is not possible after invoking rcu_irq_enter() without undoing it. +* +* For user mode entries irqentry_enter_from_user_mode() is invoked to +* establish the proper context for NOHZ_FULL. Otherwise scheduling on exit +* would not be possible. +* +* Returns: An opaque object that must be passed to idtentry_exit() +*/ +irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs); + +/** +* irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt +* +* Conditional reschedule with additional sanity checks. +*/ +void irqentry_exit_cond_resched (void ); + +/** +* irqentry_exit - Handle return from exception that used irqentry_enter() +* @regs: Pointer to pt_regs (exception entry regs) +* @state: Return value from matching call to irqentry_enter() +* +* Depending on the return target (kernel/user) this runs the necessary +* preemption and work checks if possible and reguired and returns to +* the caller with interrupts disabled and no further work pending. +* +* This is the last action before returning to the low level ASM code which +* just needs to return to the appropriate context. +* +* Counterpart to irqentry_enter(). +*/ +void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state); +#endif
kernel/entry/common.c +117 -0
120 lines not on this page · raw message
PATCH v5 5/15 entry: Provide infrastructure for work before transitioning to guest mode
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Entering a guest is similar to exiting to user space. Pending work like handling signals, rescheduling, task work etc. needs to be handled before that.
Provide generic infrastructure to avoid duplication of the same handling code all over the place.
The transfer to guest mode handling is different from the exit to usermode handling, e.g. vs. rseq and live patching, so a separate function is used.
The initial list of work items handled is:
TIF_SIGPENDING, TIF_NEED_RESCHED, TIF_NOTIFY_RESUME
Architecture specific TIF flags can be added via defines in the architecture specific include files.
The calling convention is also different from the syscall/interrupt entry functions as KVM invokes this from the outer vcpu_run() loop with interrupts and preemption enabled. To prevent missing a pending work item it invokes a check for pending TIF work from interrupt disabled code right before transitioning to guest mode. The lockdep, RCU and tracing state handling is also done directly around the switch to and from guest mode.
Signed-off-by Thomas Gleixner <[email protected]>
---
V5: Rename exit -> xfer (Sean)
V3: Reworked and simplified version adopted to recent X86 and KVM changes
V2: Moved KVM specific functions to kvm (Paolo)
Added lockdep assert (Andy)
Dropped live patching from enter guest mode work (Miroslav)
--- include/linux/entry-kvm.h | 80 ++++++++++++++++++++++++++++++++++++++++++++++ include/linux/kvm_host.h | 8 ++++ kernel/entry/Makefile | 3 + kernel/entry/kvm.c | 51 +++++++++++++++++++++++++++++ virt/kvm/Kconfig | 3 + 5 files changed, 144 insertions(+), 1 deletion(-)
include/linux/entry-kvm.h new file +80 -0
@@ -0,0 +1,80 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __LINUX_ENTRYKVM_H +#define __LINUX_ENTRYKVM_H + +#include <linux/entry-common.h> + +/* Transfer to guest mode work */ +#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK + +#ifndef ARCH_XFER_TO_GUEST_MODE_WORK +# define ARCH_XFER_TO_GUEST_MODE_WORK (0) +#endif + +#define XFER_TO_GUEST_MODE_WORK \ + (_TIF_NEED_RESCHED | _TIF_SIGPENDING | \ + _TIF_NOTIFY_RESUME | ARCH_XFER_TO_GUEST_MODE_WORK) + +struct kvm_vcpu ; + +/** +* arch_xfer_to_guest_mode_work - Architecture specific xfer to guest mode +* work function. +* @vcpu: Pointer to current's VCPU data +* @ti_work: Cached TIF flags gathered in xfer_to_guest_mode() +* +* Invoked from xfer_to_guest_mode_work(). Defaults to NOOP. Can be +* replaced by architecture specific code. +*/ +static inline int arch_xfer_to_guest_mode_work (struct kvm_vcpu *vcpu, +unsigned long ti_work); + +#ifndef arch_xfer_to_guest_mode_work +static inline int arch_xfer_to_guest_mode_work (struct kvm_vcpu *vcpu, +unsigned long ti_work) +{ +return 0 ; +} +#endif + +/** +* xfer_to_guest_mode - Check and handle pending work which needs to be +* handled before returning to guest mode +* @vcpu: Pointer to current's VCPU data +* +* Returns: 0 or an error code +*/ +int xfer_to_guest_mode (struct kvm_vcpu *vcpu); + +/** +* __xfer_to_guest_mode_work_pending - Check if work is pending +* +* Returns: True if work pending, False otherwise. +* +* Bare variant of xfer_to_guest_mode_work_pending(). Can be called from +* interrupt enabled code for racy quick checks with care. +*/ +static inline bool __xfer_to_guest_mode_work_pending (void ) +{ +unsigned long ti_work =READ_ONCE (current_thread_info ()->flags); + +return !!(ti_work &XFER_TO_GUEST_MODE_WORK ); +} + +/** +* xfer_to_guest_mode_work_pending - Check if work is pending which needs to be +* handled before returning to guest mode +* +* Returns: True if work pending, False otherwise. +* +* Has to be invoked with interrupts disabled before the transition to +* guest mode. +*/ +static inline bool xfer_to_guest_mode_work_pending (void ) +{ +lockdep_assert_irqs_disabled (); +return __xfer_to_guest_mode_work_pending (); +} +#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ + +#endif
include/linux/kvm_host.h +8 -0
@@ -1439,4 +1439,12 @@ int kvm_vm_create_worker_thread(struct k uintptr_t data, const char *name, struct task_struct **thread_ptr); +#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK +static inline void kvm_handle_signal_exit (struct kvm_vcpu *vcpu) +{ + vcpu->run->exit_reason =KVM_EXIT_INTR ; + vcpu->stat.signal_exits++; +} +#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ +#endif
kernel/entry/Makefile +2 -1
@@ -9,4 +9,5 @@ KCOV_INSTRUMENT := n CFLAGS_REMOVE_common.o = -fstack-protector -fstack-protector-strong CFLAGS_common.o += -fno-stack-protector -obj-$(CONFIG_GENERIC_ENTRY) += common.o +obj-$(CONFIG_GENERIC_ENTRY) += common.o +obj-$(CONFIG_KVM_XFER_TO_GUEST_WORK) += kvm.o
kernel/entry/kvm.c new file +51 -0
51 lines not on this page · raw message
virt/kvm/Kconfig +3 -0
@@ -60,3 +60,6 @@ config HAVE_KVM_VCPU_RUN_PID_CHANGE config HAVE_KVM_NO_POLL bool + +config KVM_XFER_TO_GUEST_WORK + bool
[tip: core/entry] entry: Provide infrastructure for work before transitioning to guest mode
The following commit has been merged into the core/entry branch of tip:
The following commit has been merged into the core/entry branch of tip:
Commit-ID: 935ace2fb5cc49ae88bd1f1735ddc51cdc2ebfb3 Gitweb: https://git.kernel.org/tip/935ace2fb5cc49ae88bd1f1735ddc51cdc2ebfb3 Author: Thomas Gleixner <[email protected]> AuthorDate: Wed, 22 Jul 2020 23:59:59 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:03:42 +02:00
entry: Provide infrastructure for work before transitioning to guest mode
Entering a guest is similar to exiting to user space. Pending work like handling signals, rescheduling, task work etc. needs to be handled before that.
Provide generic infrastructure to avoid duplication of the same handling code all over the place.
The transfer to guest mode handling is different from the exit to usermode handling, e.g. vs. rseq and live patching, so a separate function is used.
The initial list of work items handled is:
TIF_SIGPENDING, TIF_NEED_RESCHED, TIF_NOTIFY_RESUME
Architecture specific TIF flags can be added via defines in the architecture specific include files.
The calling convention is also different from the syscall/interrupt entry functions as KVM invokes this from the outer vcpu_run() loop with interrupts and preemption enabled. To prevent missing a pending work item it invokes a check for pending TIF work from interrupt disabled code right before transitioning to guest mode. The lockdep, RCU and tracing state handling is also done directly around the switch to and from guest mode.
Signed-off-by Thomas Gleixner <[email protected]>
--- include/linux/entry-kvm.h | 80 ++++++++++++++++++++++++++++++++++++++- include/linux/kvm_host.h | 8 ++++- kernel/entry/Makefile | 3 +- kernel/entry/kvm.c | 51 ++++++++++++++++++++++++- virt/kvm/Kconfig | 3 +- 5 files changed, 144 insertions(+), 1 deletion(-) create mode 100644 include/linux/entry-kvm.h create mode 100644 kernel/entry/kvm.c
include/linux/entry-kvm.h new file +80 -0
@@ -0,0 +1,80 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __LINUX_ENTRYKVM_H +#define __LINUX_ENTRYKVM_H + +#include <linux/entry-common.h> + +/* Transfer to guest mode work */ +#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK + +#ifndef ARCH_XFER_TO_GUEST_MODE_WORK +# define ARCH_XFER_TO_GUEST_MODE_WORK (0) +#endif + +#define XFER_TO_GUEST_MODE_WORK \ + (_TIF_NEED_RESCHED | _TIF_SIGPENDING | \ + _TIF_NOTIFY_RESUME | ARCH_XFER_TO_GUEST_MODE_WORK) + +struct kvm_vcpu ; + +/** +* arch_xfer_to_guest_mode_handle_work - Architecture specific xfer to guest +* mode work handling function. +* @vcpu: Pointer to current's VCPU data +* @ti_work: Cached TIF flags gathered in xfer_to_guest_mode_handle_work() +* +* Invoked from xfer_to_guest_mode_handle_work(). Defaults to NOOP. Can be +* replaced by architecture specific code. +*/ +static inline int arch_xfer_to_guest_mode_handle_work (struct kvm_vcpu *vcpu, +unsigned long ti_work); + +#ifndef arch_xfer_to_guest_mode_work +static inline int arch_xfer_to_guest_mode_handle_work (struct kvm_vcpu *vcpu, +unsigned long ti_work) +{ +return 0 ; +} +#endif + +/** +* xfer_to_guest_mode_handle_work - Check and handle pending work which needs +* to be handled before going to guest mode +* @vcpu: Pointer to current's VCPU data +* +* Returns: 0 or an error code +*/ +int xfer_to_guest_mode_handle_work (struct kvm_vcpu *vcpu); + +/** +* __xfer_to_guest_mode_work_pending - Check if work is pending +* +* Returns: True if work pending, False otherwise. +* +* Bare variant of xfer_to_guest_mode_work_pending(). Can be called from +* interrupt enabled code for racy quick checks with care. +*/ +static inline bool __xfer_to_guest_mode_work_pending (void ) +{ +unsigned long ti_work =READ_ONCE (current_thread_info ()->flags); + +return !!(ti_work &XFER_TO_GUEST_MODE_WORK ); +} + +/** +* xfer_to_guest_mode_work_pending - Check if work is pending which needs to be +* handled before returning to guest mode +* +* Returns: True if work pending, False otherwise. +* +* Has to be invoked with interrupts disabled before the transition to +* guest mode. +*/ +static inline bool xfer_to_guest_mode_work_pending (void ) +{ +lockdep_assert_irqs_disabled (); +return __xfer_to_guest_mode_work_pending (); +} +#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ + +#endif
include/linux/kvm_host.h +8 -0
@@ -1439,4 +1439,12 @@ int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn, uintptr_t data, const char *name, struct task_struct **thread_ptr); +#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK +static inline void kvm_handle_signal_exit (struct kvm_vcpu *vcpu) +{ + vcpu->run->exit_reason =KVM_EXIT_INTR ; + vcpu->stat.signal_exits++; +} +#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ +#endif
kernel/entry/Makefile +2 -1
@@ -9,4 +9,5 @@ KCOV_INSTRUMENT := n CFLAGS_REMOVE_common.o = -fstack-protector -fstack-protector-strong CFLAGS_common.o += -fno-stack-protector -obj-$(CONFIG_GENERIC_ENTRY) += common.o +obj-$(CONFIG_GENERIC_ENTRY) += common.o +obj-$(CONFIG_KVM_XFER_TO_GUEST_WORK) += kvm.o
kernel/entry/kvm.c new file +51 -0
51 lines not on this page · raw message
virt/kvm/Kconfig +3 -0
@@ -60,3 +60,6 @@ config HAVE_KVM_VCPU_RUN_PID_CHANGE config HAVE_KVM_NO_POLL bool + +config KVM_XFER_TO_GUEST_WORK + bool
Re: [patch V5 05/15] entry: Provide infrastructure for work before transitioning to guest mode
In reply to Thomas GleixnerQuoted 19 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Entering a guest is similar to exiting to user space. Pending work like handling signals, rescheduling, task work etc. needs to be handled before that.
Provide generic infrastructure to avoid duplication of the same handling code all over the place.
The transfer to guest mode handling is different from the exit to usermode handling, e.g. vs. rseq and live patching, so a separate function is used.
The initial list of work items handled is:
TIF_SIGPENDING,TIF_NEED_RESCHED,TIF_NOTIFY_RESUMEArchitecture specific TIF flags can be added via defines in the architecture specific include files.
Quoted 8 lines
The calling convention is also different from the syscall/interrupt entry functions as KVM invokes this from the outer
vcpu_run()loop with interrupts and preemption enabled. To prevent missing a pending work item it invokes a check for pending TIF work from interrupt disabled code right before transitioning to guest mode. The lockdep, RCU and tracing state handling is also done directly around the switch to and from guest mode.Signed-off-by Thomas Gleixner <[email protected]>
SR-IOV will start trigger a warning below in this commit,
A reproducer:
# echo 1 > /sys/class/net/eno58/device/sriov_numvfs (bnx2x) # git clone https://gitlab.com/cailca/linux-mm # cd linux-mm; make # ./random -x 0-100 -k0000:02:09.0
https://gitlab.com/cailca/linux-mm/-/blob/master/random.c#L1247
(x86.config is also included in the repo.)
[ 765.409027] ------------[ cut here ]------------ [ 765.434611] WARNING: CPU: 13 PID: 3377 at include/linux/entry-kvm.h:75 kvm_arch_vcpu_ioctl_run+0xb52/0x1320 [kvm] [ 765.487229] Modules linked in: vfio_pci vfio_virqfd vfio_iommu_type1 vfio loop nls_ascii nls_cp437 vfat fat [ 766.118568] CPU: 13 PID: 3377 Comm: qemu-kvm Not tainted 5.8.0-rc7-next-20200729 #2 [ 766.147011] ? kthread_create_worker_on_cpu+0xc0/0xc0 [ 766.147016] ret_from_fork+0x22/0x30 [ 766.782999] Hardware name: HP ProLiant BL660c Gen9, BIOS I38 10/17/2018 [ 766.817799] RIP: 0010:kvm_arch_vcpu_ioctl_run+0xb52/0x1320 [kvm] [ 766.850054] Code: e8 03 0f b6 04 18 84 c0 74 06 0f 8e 4a 03 00 00 41 c6 85 50 2d 00 00 00 e9 24 f8 ff ff 4c 89 ef e8 93 a8 02 00 e9 3d f8 ff ff <0f> 0b e9 f2 f8 ff ff 48 81 c6 c0 01 00 00 4c 89 ef e8 18 4c fe ff [ 766.942485] RSP: 0018:ffffc90007017b58 EFLAGS: 00010202 [ 766.970899] RAX: 0000000000000001 RBX: dffffc0000000000 RCX: ffffffffc0f0ef8a [ 767.008488] RDX: 0000000000000000 RSI: 0000000000000001 RDI: ffff88956a3821a0 [ 767.045726] RBP: ffffc90007017bb8 R08: ffffed1269290010 R09: ffffed1269290010 [ 767.083242] R10: ffff88934948007f R11: ffffed126929000f R12: ffff889349480424 [ 767.120809] R13: ffff889349480040 R14: ffff88934948006c R15: ffff88980e2da000 [ 767.160531] FS: 00007f12b0e98700(0000) GS:ffff88985fa40000(0000) knlGS:0000000000000000 [ 767.203199] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 767.235083] CR2: 0000000000000000 CR3: 000000120d186002 CR4: 00000000001726e0 [ 767.272303] Call Trace: [ 767.272303] Call Trace: [ 767.286947] kvm_vcpu_ioctl+0x3e9/0xb30 [kvm] [ 767.311162] ? kvm_vcpu_block+0xd30/0xd30 [kvm] [ 767.335281] ? find_held_lock+0x33/0x1c0 [ 767.357492] ? __fget_files+0x1cb/0x300 [ 767.378687] ? lock_downgrade+0x730/0x730 [ 767.401410] ? rcu_read_lock_held+0xaa/0xc0 [ 767.424228] ? rcu_read_lock_sched_held+0xd0/0xd0 [ 767.450078] ? find_held_lock+0x33/0x1c0 [ 767.471876] ? __fget_files+0x1e5/0x300 [ 767.493841] __x64_sys_ioctl+0x315/0x102f [ 767.516579] ? generic_block_fiemap+0x60/0x60 [ 767.540678] ? syscall_enter_from_user_mode+0x1b/0x210 [ 767.568612] ? rcu_read_lock_sched_held+0xaa/0xd0 [ 767.593950] ? lockdep_hardirqs_on_prepare+0x33e/0x4e0 [ 767.621355] ? syscall_enter_from_user_mode+0x20/0x210 [ 767.649283] ? trace_hardirqs_on+0x20/0x1b5 [ 767.673770] do_syscall_64+0x33/0x40 [ 767.694699] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 767.723392] RIP: 0033:0x7f12b999a87b [ 767.744252] Code: 0f 1e fa 48 8b 05 0d 96 2c 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d dd 95 2c 00 f7 d8 64 89 8 [ 767.837418] RSP: 002b:00007f12b0e97678 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 767.877074] RAX: ffffffffffffffda RBX: 000055712c857c60 RCX: 00007f12b999a87b [ 767.914242] RDX: 0000000000000000 RSI: 000000000000ae80 RDI: 0000000000000014 [ 767.951945] RBP: 000055712c857cfb R08: 000055712ac4fad0 R09: 000055712b3f2080 [ 767.989036] R10: 0000000000000000 R11: 0000000000000246 R12: 000055712ac38100 [ 768.025851] R13: 00007ffd996e9edf R14: 00007f12becc5000 R15: 000055712c857c60 [ 768.063363] irq event stamp: 5257 [ 768.082559] hardirqs last enabled at (5273): [<ffffffffa1800ce2>] asm_sysvec_call_function_single+0x12/0x20 [ 768.132704] hardirqs last disabled at (5290): [<ffffffffa179ae3d>] irqentry_enter+0x1d/0x50 [ 768.176563] softirqs last enabled at (5108): [<ffffffffa1a0070f>] __do_softirq+0x70f/0xa9f [ 768.221270] softirqs last disabled at (5093): [<ffffffffa1800ec2>] asm_call_on_stack+0x12/0x20 [ 768.267273] ---[ end trace 8730450ad8cfee9f ]---
Quoted 187 lines
---
V5: Rename exit -> xfer (Sean)V3: Reworked and simplified version adopted to recent X86 and KVM changes
V2: Moved KVM specific functions to kvm (Paolo)
Added lockdep assert (Andy)
Dropped live patching from enter guest mode work (Miroslav)--- include/linux/entry-kvm.h | 80 ++++++++++++++++++++++++++++++++++++++++++++++ include/linux/kvm_host.h | 8 ++++ kernel/entry/Makefile | 3 + kernel/entry/kvm.c | 51 +++++++++++++++++++++++++++++ virt/kvm/Kconfig | 3 + 5 files changed, 144 insertions(+), 1 deletion(-)include/linux/entry-kvm.h new file +80
-0@@ -0,0 +1,80 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __LINUX_ENTRYKVM_H +#define __LINUX_ENTRYKVM_H + +#include <linux/entry-common.h> + +/* Transfer to guest mode work */ +#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK + +#ifndef ARCH_XFER_TO_GUEST_MODE_WORK +# define ARCH_XFER_TO_GUEST_MODE_WORK (0) +#endif + +#define XFER_TO_GUEST_MODE_WORK \ + (_TIF_NEED_RESCHED | _TIF_SIGPENDING | \ + _TIF_NOTIFY_RESUME | ARCH_XFER_TO_GUEST_MODE_WORK) + +struct kvm_vcpu ; + +/** +* arch_xfer_to_guest_mode_work - Architecture specific xfer to guest mode +* work function. +* @vcpu: Pointer to current's VCPU data +* @ti_work: Cached TIF flags gathered in xfer_to_guest_mode() +* +* Invoked from xfer_to_guest_mode_work(). Defaults to NOOP. Can be +* replaced by architecture specific code. +*/ +static inline int arch_xfer_to_guest_mode_work (struct kvm_vcpu *vcpu, +unsigned long ti_work); + +#ifndef arch_xfer_to_guest_mode_work +static inline int arch_xfer_to_guest_mode_work (struct kvm_vcpu *vcpu, +unsigned long ti_work) +{ +return 0 ; +} +#endif + +/** +* xfer_to_guest_mode - Check and handle pending work which needs to be +* handled before returning to guest mode +* @vcpu: Pointer to current's VCPU data +* +* Returns: 0 or an error code +*/ +int xfer_to_guest_mode (struct kvm_vcpu *vcpu); + +/** +* __xfer_to_guest_mode_work_pending - Check if work is pending +* +* Returns: True if work pending, False otherwise. +* +* Bare variant of xfer_to_guest_mode_work_pending(). Can be called from +* interrupt enabled code for racy quick checks with care. +*/ +static inline bool __xfer_to_guest_mode_work_pending (void ) +{ +unsigned long ti_work =READ_ONCE (current_thread_info ()->flags); + +return !!(ti_work &XFER_TO_GUEST_MODE_WORK ); +} + +/** +* xfer_to_guest_mode_work_pending - Check if work is pending which needs to be +* handled before returning to guest mode +* +* Returns: True if work pending, False otherwise. +* +* Has to be invoked with interrupts disabled before the transition to +* guest mode. +*/ +static inline bool xfer_to_guest_mode_work_pending (void ) +{ +lockdep_assert_irqs_disabled (); +return __xfer_to_guest_mode_work_pending (); +} +#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ + +#endif include/linux/kvm_host.h +8
-0@@ -1439,4 +1439,12 @@ int kvm_vm_create_worker_thread(struct k uintptr_t data, const char *name, struct task_struct **thread_ptr); +#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK +static inline void kvm_handle_signal_exit (struct kvm_vcpu *vcpu) +{ + vcpu->run->exit_reason =KVM_EXIT_INTR ; + vcpu->stat.signal_exits++; +} +#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ +#endif kernel/entry/Makefile +2
-1@@ -9,4 +9,5 @@ KCOV_INSTRUMENT := n CFLAGS_REMOVE_common.o = -fstack-protector -fstack-protector-strong CFLAGS_common.o += -fno-stack-protector -obj-$(CONFIG_GENERIC_ENTRY) += common.o +obj-$(CONFIG_GENERIC_ENTRY) += common.o +obj-$(CONFIG_KVM_XFER_TO_GUEST_WORK) += kvm.okernel/entry/kvm.c new file +51
-0@@ -0,0 +1,51 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include <linux/entry-kvm.h> +#include <linux/kvm_host.h> + +static int xfer_to_guest_mode_work (struct kvm_vcpu *vcpu,unsigned long ti_work) +{ +do { +int ret; + +if (ti_work & _TIF_SIGPENDING) { +kvm_handle_signal_exit (vcpu); +return -EINTR ; + } + +if (ti_work & _TIF_NEED_RESCHED) +schedule (); + +if (ti_work & _TIF_NOTIFY_RESUME) { +clear_thread_flag (TIF_NOTIFY_RESUME ); +tracehook_notify_resume (NULL ); + } + + ret =arch_xfer_to_guest_mode_work (vcpu, ti_work); +if (ret) +return ret; + + ti_work =READ_ONCE (current_thread_info ()->flags); + }while (ti_work &XFER_TO_GUEST_MODE_WORK ||need_resched ()); +return 0 ; +} + +int xfer_to_guest_mode (struct kvm_vcpu *vcpu) +{ +unsigned long ti_work; + +/* +* This is invoked from the outer guest loop with interrupts and +* preemption enabled. +* +* KVM invokes xfer_to_guest_mode_work_pending() with interrupts +* disabled in the inner loop before going into guest mode. No need +* to disable interrupts here. +*/ + ti_work =READ_ONCE (current_thread_info ()->flags); +if (!(ti_work &XFER_TO_GUEST_MODE_WORK )) +return 0 ; + +return xfer_to_guest_mode_work (vcpu, ti_work); +} +EXPORT_SYMBOL_GPL (xfer_to_guest_mode);virt/kvm/Kconfig +3
-0@@ -60,3 +60,6 @@ config HAVE_KVM_VCPU_RUN_PID_CHANGE config HAVE_KVM_NO_POLL bool + +config KVM_XFER_TO_GUEST_WORK + bool
Re: [patch V5 05/15] entry: Provide infrastructure for work before transitioning to guest mode
Quoted 4 lines · Qian Cai
On Wed, Jul 22, 2020 at 11:59:59PM +0200, Thomas Gleixner wrote:
SR-IOV will start trigger a warning below in this commit,[ 765.434611] WARNING: CPU: 13 PID: 3377 at include/linux/entry-kvm.h:75 kvm_arch_vcpu_ioctl_run+0xb52/0x1320 [kvm]
Yes, I'm a moron. Fixed it locally and failed to transfer the fixup when merging it. Fix below.
Quoted 2 lines
[ 768.221270] softirqs last disabled at (5093): [<ffffffffa1800ec2>] asm_call_on_stack+0x12/0x20 [ 768.267273] ---[ end trace 8730450ad8cfee9f ]---
Can you pretty please trim your replies?
Quoted 2 lines
Quoted 2 lines
---
V5: Rename exit -> xfer (Sean)
<removed 200 lines of useless information>
Thanks,
tglx
---
arch/x86/kvm/x86.c +1 -1
@@ -8682,7 +8682,7 @@ static int vcpu_run(struct kvm_vcpu *vcpu)break ; } -if (xfer_to_guest_mode_work_pending ()) { +if (__xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); r =xfer_to_guest_mode_handle_work (vcpu);if (r)
[tip: x86/entry] x86/kvm: Use __xfer_to_guest_mode_work_pending() in kvm_run_vcpu()
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: f3020b8891b890b48d9e1a83241e3cce518427c1 Gitweb: https://git.kernel.org/tip/f3020b8891b890b48d9e1a83241e3cce518427c1 Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 30 Jul 2020 09:19:01 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Thu, 30 Jul 2020 12:31:47 +02:00
x86/kvm: Use __xfer_to_guest_mode_work_pending() in kvm_run_vcpu()
The comments explicitely explain that the work flags check and handling in kvm_run_vcpu() is done with preemption and interrupts enabled as KVM invokes the check again right before entering guest mode with interrupts disabled which guarantees that the work flags are observed and handled before VMENTER.
Nevertheless the flag pending check in kvm_run_vcpu() uses the helper variant which requires interrupts to be disabled triggering an instant lockdep splat. This was caught in testing before and then not fixed up in the patch before applying. :(
Use the relaxed and intentionally racy __xfer_to_guest_mode_work_pending() instead.
Fixes 72c3c0fe54a3 ("x86/kvm: Use generic xfer to guest work function")
Reported-by Qian Cai <[email protected]> writes:
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/kvm/x86.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/kvm/x86.c +1 -1
@@ -8682,7 +8682,7 @@ static int vcpu_run(struct kvm_vcpu *vcpu)break ; } -if (xfer_to_guest_mode_work_pending ()) { +if (__xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); r =xfer_to_guest_mode_handle_work (vcpu);if (r)
PATCH v5 6/15 x86/entry: Consolidate check_user_regs()
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
The user register sanity check is sprinkled all over the place. Move it into enter_from_user_mode().
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/entry/common.c | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-)
arch/x86/entry/common.c +9 -15
@@ -82,10 +82,11 @@ static noinstr void check_user_regs(stru* 2) Invoke context tracking if enabled to reactivate RCU * 3) Trace interrupts off state */ -static noinstr void enter_from_user_mode(void) +static noinstr void enter_from_user_mode(struct pt_regs *regs) {enum ctx_state state =ct_state (); +check_user_regs (regs);lockdep_hardirqs_off (CALLER_ADDR0 );user_exit_irqoff ();
@@ -95,8 +96,9 @@ static noinstr void enter_from_user_modeinstrumentation_end (); }#else -static __always_inline void enter_from_user_mode(void) +static __always_inline void enter_from_user_mode(struct pt_regs *regs) { +check_user_regs (regs);lockdep_hardirqs_off (CALLER_ADDR0 );instrumentation_begin ();trace_hardirqs_off_finish ();
@@ -369,9 +371,7 @@ static void __syscall_return_slowpath(st {struct thread_info *ti; -check_user_regs (regs); - -enter_from_user_mode (); +enter_from_user_mode (regs);instrumentation_begin ();local_irq_enable ();
@@ -434,9 +434,7 @@ static void do_syscall_32_irqs_on(struct/* Handles int $0x80 */ __visible noinstr void do_int80_syscall_32(struct pt_regs *regs) { -check_user_regs (regs); - -enter_from_user_mode (); +enter_from_user_mode (regs);instrumentation_begin ();local_irq_enable ();
@@ -487,8 +485,6 @@ static bool __do_fast_syscall_32(struct 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.
@@ -496,7 +492,7 @@ static bool __do_fast_syscall_32(struct*/ regs->ip = landing_pad; -enter_from_user_mode (); +enter_from_user_mode (regs);instrumentation_begin ();local_irq_enable ();
@@ -599,8 +595,7 @@ idtentry_state_t noinstr idtentry_enter( };if (user_mode (regs)) { -check_user_regs (regs); -enter_from_user_mode (); +enter_from_user_mode (regs);return ret; }
@@ -733,8 +728,7 @@ void noinstr idtentry_exit(struct pt_reg*/ void noinstr idtentry_enter_user(struct pt_regs *regs) { -check_user_regs (regs); -enter_from_user_mode (); +enter_from_user_mode (regs); }/**
[tip: x86/entry] x86/entry: Consolidate check_user_regs()
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 8d5ea35c5e9139dbd19a3d73985d008d36c9968f Gitweb: https://git.kernel.org/tip/8d5ea35c5e9139dbd19a3d73985d008d36c9968f Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:00 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:57 +02:00
x86/entry: Consolidate check_user_regs()
The user register sanity check is sprinkled all over the place. Move it into enter_from_user_mode().
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/entry/common.c | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-)
arch/x86/entry/common.c +9 -15
@@ -82,10 +82,11 @@ static noinstr void check_user_regs(struct pt_regs *regs)* 2) Invoke context tracking if enabled to reactivate RCU * 3) Trace interrupts off state */ -static noinstr void enter_from_user_mode(void) +static noinstr void enter_from_user_mode(struct pt_regs *regs) {enum ctx_state state =ct_state (); +check_user_regs (regs);lockdep_hardirqs_off (CALLER_ADDR0 );user_exit_irqoff ();
@@ -95,8 +96,9 @@ static noinstr void enter_from_user_mode(void)instrumentation_end (); }#else -static __always_inline void enter_from_user_mode(void) +static __always_inline void enter_from_user_mode(struct pt_regs *regs) { +check_user_regs (regs);lockdep_hardirqs_off (CALLER_ADDR0 );instrumentation_begin ();trace_hardirqs_off_finish ();
@@ -369,9 +371,7 @@ __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 (); +enter_from_user_mode (regs);instrumentation_begin ();local_irq_enable ();
@@ -434,9 +434,7 @@ 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 (); +enter_from_user_mode (regs);instrumentation_begin ();local_irq_enable ();
@@ -487,8 +485,6 @@ __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.
@@ -496,7 +492,7 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)*/ regs->ip = landing_pad; -enter_from_user_mode (); +enter_from_user_mode (regs);instrumentation_begin ();local_irq_enable ();
@@ -599,8 +595,7 @@ idtentry_state_t noinstr idtentry_enter(struct pt_regs *regs) };if (user_mode (regs)) { -check_user_regs (regs); -enter_from_user_mode (); +enter_from_user_mode (regs);return ret; }
@@ -733,8 +728,7 @@ void noinstr idtentry_exit(struct pt_regs *regs, idtentry_state_t state)*/ void noinstr idtentry_enter_user(struct pt_regs *regs) { -check_user_regs (regs); -enter_from_user_mode (); +enter_from_user_mode (regs); }/**
PATCH v5 7/15 x86/entry: Consolidate 32/64 bit syscall entry
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
64bit and 32bit entry code have the same open coded syscall entry handling after the bitwidth specific bits.
Move it to a helper function and share the code.
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/entry/common.c | 93 +++++++++++++++++++++--------------------------- 1 file changed, 41 insertions(+), 52 deletions(-)
arch/x86/entry/common.c +41 -52
@@ -366,8 +366,7 @@ static void __syscall_return_slowpath(stexit_to_user_mode (); } -#ifdef CONFIG_X86_64 -__visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) +static noinstr long syscall_enter (struct pt_regs *regs,unsigned long nr) {struct thread_info *ti;
@@ -379,6 +378,16 @@ static void __syscall_return_slowpath(stif (READ_ONCE (ti->flags) & _TIF_WORK_SYSCALL_ENTRY) nr =syscall_trace_enter (regs); +instrumentation_end (); +return nr; +} + +#ifdef CONFIG_X86_64 +__visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) +{ + nr =syscall_enter (regs, nr); + +instrumentation_begin ();if (likely (nr < NR_syscalls)) { nr =array_index_nospec (nr, NR_syscalls); regs->ax = sys_call_table[nr](regs);
@@ -390,64 +399,53 @@ static void __syscall_return_slowpath(st regs->ax = x32_sys_call_table[nr](regs); #endif } -__syscall_return_slowpath (regs); -instrumentation_end (); -exit_to_user_mode (); +syscall_return_slowpath (regs); }#endif #if defined(CONFIG_X86_32 ) || defined(CONFIG_IA32_EMULATION ) +static __always_inline unsigned int syscall_32_enter(struct pt_regs *regs) +{ +if (IS_ENABLED (CONFIG_IA32_EMULATION )) +current_thread_info ()->status |=TS_COMPAT ; +/* +* Subtlety here: if ptrace pokes something larger than 2^32-1 into +* orig_ax, the unsigned int return value truncates it. This may +* or may not be necessary, but it matches the old asm behavior. +*/ +return syscall_enter (regs, (unsigned int )regs->orig_ax); +} +/* -* Does a 32-bit syscall. Called with IRQs on in CONTEXT_KERNEL. Does -* all entry and exit work and returns with IRQs off. This function is -* extremely hot in workloads that use it, and it's usually called from -* do_fast_syscall_32, so forcibly inline it to improve performance. +* Invoke a 32-bit syscall. Called with IRQs on in CONTEXT_KERNEL. */ -static void do_syscall_32_irqs_on (struct pt_regs *regs) +static __always_inline void do_syscall_32_irqs_on(struct pt_regs *regs, +unsigned int nr) { -struct thread_info *ti =current_thread_info (); -unsigned int nr = (unsigned int )regs->orig_ax; - -#ifdef CONFIG_IA32_EMULATION - ti->status |=TS_COMPAT ; -#endif - -if (READ_ONCE (ti->flags) & _TIF_WORK_SYSCALL_ENTRY) { -/* -* Subtlety here: if ptrace pokes something larger than -* 2^32-1 into orig_ax, this truncates it. This may or -* may not be necessary, but it matches the old asm -* behavior. -*/ - nr =syscall_trace_enter (regs); - } -if (likely (nr < IA32_NR_syscalls)) { +instrumentation_begin (); nr =array_index_nospec (nr, IA32_NR_syscalls); regs->ax = ia32_sys_call_table[nr](regs); +instrumentation_end (); } - -__syscall_return_slowpath (regs); }/* Handles int $0x80 */ __visible noinstr void do_int80_syscall_32(struct pt_regs *regs) { -enter_from_user_mode (regs); -instrumentation_begin (); - -local_irq_enable (); -do_syscall_32_irqs_on (regs); +unsigned int nr =syscall_32_enter (regs); -instrumentation_end (); -exit_to_user_mode (); +do_syscall_32_irqs_on (regs, nr); +syscall_return_slowpath (regs); } -static bool __do_fast_syscall_32 (struct pt_regs *regs) +static noinstr bool __do_fast_syscall_32(struct pt_regs *regs) { +unsigned int nr =syscall_32_enter (regs);int res; +instrumentation_begin ();/* Fetch EBP from where the vDSO stashed it. */ if (IS_ENABLED (CONFIG_X86_64 )) {/*
@@ -460,17 +458,18 @@ static bool __do_fast_syscall_32(struct res =get_user (*(u32 *)®s->bp, (u32 __user __force *)(unsigned long )(u32 )regs->sp); } +instrumentation_end ();if (res) {/* User code screwed up. */ regs->ax = -EFAULT ; -local_irq_disable (); -__prepare_exit_to_usermode (regs); +syscall_return_slowpath (regs);return false; }/* Now this is just like a normal syscall. */ -do_syscall_32_irqs_on (regs); +do_syscall_32_irqs_on (regs, nr); +syscall_return_slowpath (regs);return true; }
@@ -483,7 +482,6 @@ static bool __do_fast_syscall_32(struct*/ unsigned long landing_pad = (unsigned long )current->mm->context.vdso + vdso_image_32.sym_int80_landing_pad; -bool success;/* * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
@@ -492,17 +490,8 @@ static bool __do_fast_syscall_32(struct*/ regs->ip = landing_pad; -enter_from_user_mode (regs); -instrumentation_begin (); - -local_irq_enable (); - success =__do_fast_syscall_32 (regs); - -instrumentation_end (); -exit_to_user_mode (); - -/* If it failed, keep it simple: use IRET. */ -if (!success) +/* Invoke the syscall. If it failed, keep it simple: use IRET. */ +if (!__do_fast_syscall_32 (regs))return 0 ;#ifdef CONFIG_X86_64
[tip: x86/entry] x86/entry: Consolidate 32/64 bit syscall entry
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 0b085e68f4072024ecaa3889aeeaab5f6c8eba5c Gitweb: https://git.kernel.org/tip/0b085e68f4072024ecaa3889aeeaab5f6c8eba5c Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:01 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:58 +02:00
x86/entry: Consolidate 32/64 bit syscall entry
64bit and 32bit entry code have the same open coded syscall entry handling after the bitwidth specific bits.
Move it to a helper function and share the code.
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/entry/common.c | 93 +++++++++++++++++----------------------- 1 file changed, 41 insertions(+), 52 deletions(-)
arch/x86/entry/common.c +41 -52
@@ -366,8 +366,7 @@ __visible noinstr void syscall_return_slowpath(struct pt_regs *regs)exit_to_user_mode (); } -#ifdef CONFIG_X86_64 -__visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) +static noinstr long syscall_enter (struct pt_regs *regs,unsigned long nr) {struct thread_info *ti;
@@ -379,6 +378,16 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs)if (READ_ONCE (ti->flags) & _TIF_WORK_SYSCALL_ENTRY) nr =syscall_trace_enter (regs); +instrumentation_end (); +return nr; +} + +#ifdef CONFIG_X86_64 +__visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) +{ + nr =syscall_enter (regs, nr); + +instrumentation_begin ();if (likely (nr < NR_syscalls)) { nr =array_index_nospec (nr, NR_syscalls); regs->ax = sys_call_table[nr](regs);
@@ -390,64 +399,53 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) regs->ax = x32_sys_call_table[nr](regs); #endif } -__syscall_return_slowpath (regs); -instrumentation_end (); -exit_to_user_mode (); +syscall_return_slowpath (regs); }#endif #if defined(CONFIG_X86_32 ) || defined(CONFIG_IA32_EMULATION ) +static __always_inline unsigned int syscall_32_enter(struct pt_regs *regs) +{ +if (IS_ENABLED (CONFIG_IA32_EMULATION )) +current_thread_info ()->status |=TS_COMPAT ; +/* +* Subtlety here: if ptrace pokes something larger than 2^32-1 into +* orig_ax, the unsigned int return value truncates it. This may +* or may not be necessary, but it matches the old asm behavior. +*/ +return syscall_enter (regs, (unsigned int )regs->orig_ax); +} +/* -* Does a 32-bit syscall. Called with IRQs on in CONTEXT_KERNEL. Does -* all entry and exit work and returns with IRQs off. This function is -* extremely hot in workloads that use it, and it's usually called from -* do_fast_syscall_32, so forcibly inline it to improve performance. +* Invoke a 32-bit syscall. Called with IRQs on in CONTEXT_KERNEL. */ -static void do_syscall_32_irqs_on (struct pt_regs *regs) +static __always_inline void do_syscall_32_irqs_on(struct pt_regs *regs, +unsigned int nr) { -struct thread_info *ti =current_thread_info (); -unsigned int nr = (unsigned int )regs->orig_ax; - -#ifdef CONFIG_IA32_EMULATION - ti->status |=TS_COMPAT ; -#endif - -if (READ_ONCE (ti->flags) & _TIF_WORK_SYSCALL_ENTRY) { -/* -* Subtlety here: if ptrace pokes something larger than -* 2^32-1 into orig_ax, this truncates it. This may or -* may not be necessary, but it matches the old asm -* behavior. -*/ - nr =syscall_trace_enter (regs); - } -if (likely (nr < IA32_NR_syscalls)) { +instrumentation_begin (); nr =array_index_nospec (nr, IA32_NR_syscalls); regs->ax = ia32_sys_call_table[nr](regs); +instrumentation_end (); } - -__syscall_return_slowpath (regs); }/* Handles int $0x80 */ __visible noinstr void do_int80_syscall_32(struct pt_regs *regs) { -enter_from_user_mode (regs); -instrumentation_begin (); +unsigned int nr =syscall_32_enter (regs); -local_irq_enable (); -do_syscall_32_irqs_on (regs); - -instrumentation_end (); -exit_to_user_mode (); +do_syscall_32_irqs_on (regs, nr); +syscall_return_slowpath (regs); } -static bool __do_fast_syscall_32 (struct pt_regs *regs) +static noinstr bool __do_fast_syscall_32(struct pt_regs *regs) { +unsigned int nr =syscall_32_enter (regs);int res; +instrumentation_begin ();/* Fetch EBP from where the vDSO stashed it. */ if (IS_ENABLED (CONFIG_X86_64 )) {/*
@@ -460,17 +458,18 @@ static bool __do_fast_syscall_32(struct pt_regs *regs) res =get_user (*(u32 *)®s->bp, (u32 __user __force *)(unsigned long )(u32 )regs->sp); } +instrumentation_end ();if (res) {/* User code screwed up. */ regs->ax = -EFAULT ; -local_irq_disable (); -__prepare_exit_to_usermode (regs); +syscall_return_slowpath (regs);return false; }/* Now this is just like a normal syscall. */ -do_syscall_32_irqs_on (regs); +do_syscall_32_irqs_on (regs, nr); +syscall_return_slowpath (regs);return true; }
@@ -483,7 +482,6 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)*/ unsigned long landing_pad = (unsigned long )current->mm->context.vdso + vdso_image_32.sym_int80_landing_pad; -bool success;/* * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
@@ -492,17 +490,8 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)*/ regs->ip = landing_pad; -enter_from_user_mode (regs); -instrumentation_begin (); - -local_irq_enable (); - success =__do_fast_syscall_32 (regs); - -instrumentation_end (); -exit_to_user_mode (); - -/* If it failed, keep it simple: use IRET. */ -if (!success) +/* Invoke the syscall. If it failed, keep it simple: use IRET. */ +if (!__do_fast_syscall_32 (regs))return 0 ;#ifdef CONFIG_X86_64
Re: [tip: x86/entry] x86/entry: Consolidate 32/64 bit syscall entry
Quoted 127 lines · On Fri, Jul 24, 2020 at 4:14 PM tip-bot2 for Thomas Gleixner…
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 0b085e68f4072024ecaa3889aeeaab5f6c8eba5c Gitweb: https://git.kernel.org/tip/0b085e68f4072024ecaa3889aeeaab5f6c8eba5c Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:01 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:58 +02:00x86/entry: Consolidate 32/64 bit syscall entry
64bit and 32bit entry code have the same open coded syscall entry handling after the bitwidth specific bits.
Move it to a helper function and share the code.
Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/entry/common.c | 93 +++++++++++++++++----------------------- 1 file changed, 41 insertions(+), 52 deletions(-)arch/x86/entry/common.c +30
-32@@ -366,8 +366,7 @@ __visible noinstr void syscall_return_slowpath(struct pt_regs *regs)exit_to_user_mode (); } -#ifdef CONFIG_X86_64 -__visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) +static noinstr long syscall_enter (struct pt_regs *regs,unsigned long nr) {struct thread_info *ti;@@ -379,6 +378,16 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs)if (READ_ONCE (ti->flags) & _TIF_WORK_SYSCALL_ENTRY) nr =syscall_trace_enter (regs); +instrumentation_end (); +return nr; +} + +#ifdef CONFIG_X86_64 +__visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) +{ + nr =syscall_enter (regs, nr); + +instrumentation_begin ();if (likely (nr < NR_syscalls)) { nr =array_index_nospec (nr, NR_syscalls); regs->ax = sys_call_table[nr](regs);@@ -390,64 +399,53 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) regs->ax = x32_sys_call_table[nr](regs); #endif } -__syscall_return_slowpath (regs); -instrumentation_end (); -exit_to_user_mode (); +syscall_return_slowpath (regs); }#endif #if defined(CONFIG_X86_32 ) || defined(CONFIG_IA32_EMULATION ) +static __always_inline unsigned int syscall_32_enter(struct pt_regs *regs) +{ +if (IS_ENABLED (CONFIG_IA32_EMULATION )) +current_thread_info ()->status |=TS_COMPAT ; +/* +* Subtlety here: if ptrace pokes something larger than 2^32-1 into +* orig_ax, the unsigned int return value truncates it. This may +* or may not be necessary, but it matches the old asm behavior. +*/ +return syscall_enter (regs, (unsigned int )regs->orig_ax); +} +/* -* Does a 32-bit syscall. Called with IRQs on in CONTEXT_KERNEL. Does -* all entry and exit work and returns with IRQs off. This function is -* extremely hot in workloads that use it, and it's usually called from -* do_fast_syscall_32, so forcibly inline it to improve performance. +* Invoke a 32-bit syscall. Called with IRQs on in CONTEXT_KERNEL. */ -static void do_syscall_32_irqs_on (struct pt_regs *regs) +static __always_inline void do_syscall_32_irqs_on(struct pt_regs *regs, +unsigned int nr) { -struct thread_info *ti =current_thread_info (); -unsigned int nr = (unsigned int )regs->orig_ax; - -#ifdef CONFIG_IA32_EMULATION - ti->status |=TS_COMPAT ; -#endif - -if (READ_ONCE (ti->flags) & _TIF_WORK_SYSCALL_ENTRY) { -/* -* Subtlety here: if ptrace pokes something larger than -* 2^32-1 into orig_ax, this truncates it. This may or -* may not be necessary, but it matches the old asm -* behavior. -*/ - nr =syscall_trace_enter (regs); - } -if (likely (nr < IA32_NR_syscalls)) { +instrumentation_begin (); nr =array_index_nospec (nr, IA32_NR_syscalls); regs->ax = ia32_sys_call_table[nr](regs); +instrumentation_end (); } - -__syscall_return_slowpath (regs); }/* Handles int $0x80 */ __visible noinstr void do_int80_syscall_32(struct pt_regs *regs) { -enter_from_user_mode (regs); -instrumentation_begin (); +unsigned int nr =syscall_32_enter (regs); -local_irq_enable ();
arch/x86/entry/common.c __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) · line 443
Quoted 10 lines
arch/x86/entry/common.c +3
-5-do_syscall_32_irqs_on (regs); - -instrumentation_end (); -exit_to_user_mode (); +do_syscall_32_irqs_on (regs, nr); +syscall_return_slowpath (regs); } -static bool __do_fast_syscall_32 (struct pt_regs *regs) +static noinstr bool __do_fast_syscall_32(struct pt_regs *regs)
Can __do_fast_syscall_32() be merged back into do_fast_syscall_32() now that both are marked noinstr?
--
Brian Gerst
Re: [tip: x86/entry] x86/entry: Consolidate 32/64 bit syscall entry
Quoted 7 lines · Brian Gerst
On Fri, Jul 24, 2020 at 4:14 PM tip-bot2 for Thomas Gleixner
arch/x86/entry/common.c __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) · line 443
Quoted 3 lines
arch/x86/entry/common.c +1
-1-static bool __do_fast_syscall_32 (struct pt_regs *regs) +static noinstr bool __do_fast_syscall_32(struct pt_regs *regs)Can
__do_fast_syscall_32()be merged back intodo_fast_syscall_32()now that both are marked noinstr?
It could.
PATCH v5 8/15 x86/entry: Move user return notifier out of loop
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Guests and user space share certain MSRs. KVM sets these MSRs to guest values once and does not set them back to user space values on every VM exit to spare the costly MSR operations.
User return notifiers ensure that these MSRs are set back to the correct values before returning to user space in exit_to_usermode_loop().
There is no reason to evaluate the TIF flag indicating that user return notifiers need to be invoked in the loop. The important point is that they are invoked before returning to user space.
Move the invocation out of the loop into the section which does the last preperatory steps before returning to user space. That section is not preemptible and runs with interrupts disabled until the actual return.
Signed-off-by Thomas Gleixner <[email protected]>
---
V4: New patch
--- arch/x86/entry/common.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
arch/x86/entry/common.c +4 -4
@@ -208,7 +208,7 @@ static long syscall_trace_enter(struct p#define EXIT_TO_USERMODE_LOOP_FLAGS \ (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \ - _TIF_NEED_RESCHED | _TIF_USER_RETURN_NOTIFY | _TIF_PATCH_PENDING) + _TIF_NEED_RESCHED | _TIF_PATCH_PENDING)static void exit_to_usermode_loop (struct pt_regs *regs,u32 cached_flags) {
@@ -242,9 +242,6 @@ static void exit_to_usermode_loop(structrseq_handle_notify_resume (NULL , regs); } -if (cached_flags & _TIF_USER_RETURN_NOTIFY) -fire_user_return_notifiers (); -/* Disable IRQs and retry */ local_irq_disable ();
@@ -273,6 +270,9 @@ static void __prepare_exit_to_usermode(s/* Reload ti->flags; we may have rescheduled above. */ cached_flags =READ_ONCE (ti->flags); +if (cached_flags & _TIF_USER_RETURN_NOTIFY) +fire_user_return_notifiers (); +if (unlikely (cached_flags & _TIF_IO_BITMAP))tss_update_io_bitmap ();
Re: [patch V5 08/15] x86/entry: Move user return notifier out of loop
Quoted 13 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Guests and user space share certain MSRs. KVM sets these MSRs to guest values once and does not set them back to user space values on every VM exit to spare the costly MSR operations.
User return notifiers ensure that these MSRs are set back to the correct values before returning to user space in
exit_to_usermode_loop().There is no reason to evaluate the TIF flag indicating that user return notifiers need to be invoked in the loop. The important point is that they are invoked before returning to user space.
Quoted 5 lines
Move the invocation out of the loop into the section which does the last preperatory steps before returning to user space. That section is not preemptible and runs with interrupts disabled until the actual return.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-and-tested-by Sean Christopherson <[email protected]>
[tip: x86/entry] x86/entry: Move user return notifier out of loop
In reply to Thomas GleixnerThe following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: a377ac1cd9d7b9ac8d546dceb3d74956fbfd443f Gitweb: https://git.kernel.org/tip/a377ac1cd9d7b9ac8d546dceb3d74956fbfd443f Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:02 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:58 +02:00
x86/entry: Move user return notifier out of loop
Guests and user space share certain MSRs. KVM sets these MSRs to guest values once and does not set them back to user space values on every VM exit to spare the costly MSR operations.
User return notifiers ensure that these MSRs are set back to the correct values before returning to user space in exit_to_usermode_loop().
There is no reason to evaluate the TIF flag indicating that user return notifiers need to be invoked in the loop. The important point is that they are invoked before returning to user space.
Move the invocation out of the loop into the section which does the last preperatory steps before returning to user space. That section is not preemptible and runs with interrupts disabled until the actual return.
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/entry/common.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
arch/x86/entry/common.c +4 -4
@@ -208,7 +208,7 @@ static long syscall_trace_enter(struct pt_regs *regs)#define EXIT_TO_USERMODE_LOOP_FLAGS \ (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \ - _TIF_NEED_RESCHED | _TIF_USER_RETURN_NOTIFY | _TIF_PATCH_PENDING) + _TIF_NEED_RESCHED | _TIF_PATCH_PENDING)static void exit_to_usermode_loop (struct pt_regs *regs,u32 cached_flags) {
@@ -242,9 +242,6 @@ static void exit_to_usermode_loop(struct pt_regs *regs, u32 cached_flags)rseq_handle_notify_resume (NULL , regs); } -if (cached_flags & _TIF_USER_RETURN_NOTIFY) -fire_user_return_notifiers (); -/* Disable IRQs and retry */ local_irq_disable ();
@@ -273,6 +270,9 @@ static void __prepare_exit_to_usermode(struct pt_regs *regs)/* Reload ti->flags; we may have rescheduled above. */ cached_flags =READ_ONCE (ti->flags); +if (cached_flags & _TIF_USER_RETURN_NOTIFY) +fire_user_return_notifiers (); +if (unlikely (cached_flags & _TIF_IO_BITMAP))tss_update_io_bitmap ();
PATCH v5 9/15 x86/ptrace: Provide pt_regs helper for entry/exit
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
As a preparatory step for moving the syscall and interrupt entry/exit handling into generic code, provide a pt_regs helper which retrieves the interrupt state from pt_regs. This is required to check whether interrupts are reenabled by return from interrupt/exception.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
---
V4: Remove the syscall nr and return value helpers as they alredy exist
--- arch/x86/include/asm/ptrace.h | 5 +++++ 1 file changed, 5 insertions(+)
arch/x86/include/asm/ptrace.h +5 -0
@@ -209,6 +209,11 @@ static inline void user_stack_pointer_se regs->sp = val; } +static __always_inline bool regs_irqs_disabled(struct pt_regs *regs) +{ +return !(regs->flags &X86_EFLAGS_IF ); +} +/* Query offset/name of register from its name/offset */ extern int regs_query_register_offset (const char *name);extern const char *regs_query_register_name (unsigned int offset);
[tip: x86/entry] x86/ptrace: Provide pt_regs helper for entry/exit
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 0bf019ea59e330770883ede4499d7f711d8c3adf Gitweb: https://git.kernel.org/tip/0bf019ea59e330770883ede4499d7f711d8c3adf Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:03 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:58 +02:00
x86/ptrace: Provide pt_regs helper for entry/exit
As a preparatory step for moving the syscall and interrupt entry/exit handling into generic code, provide a pt_regs helper which retrieves the interrupt state from pt_regs. This is required to check whether interrupts are reenabled by return from interrupt/exception.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/include/asm/ptrace.h | 5 +++++ 1 file changed, 5 insertions(+)
arch/x86/include/asm/ptrace.h +5 -0
@@ -209,6 +209,11 @@ static inline void user_stack_pointer_set(struct pt_regs *regs, regs->sp = val; } +static __always_inline bool regs_irqs_disabled(struct pt_regs *regs) +{ +return !(regs->flags &X86_EFLAGS_IF ); +} +/* Query offset/name of register from its name/offset */ extern int regs_query_register_offset (const char *name);extern const char *regs_query_register_name (unsigned int offset);
PATCH v5 10/15 x86/entry: Use generic syscall entry function
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Replace the syscall entry work handling with the generic version. Provide the necessary helper inlines to handle the real architecture specific parts, e.g. ptrace.
Use a temporary define for idtentry_enter_user which will be cleaned up seperately.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
---
V4: Drop the seccomp/audit wrappers as they are not longer required (Kees)
V3: Adapt to the noinstr changes
--- arch/x86/Kconfig | 1 arch/x86/entry/common.c | 181 +----------------------------------- arch/x86/include/asm/entry-common.h | 32 ++++++ arch/x86/include/asm/idtentry.h | 5 arch/x86/include/asm/thread_info.h | 5 5 files changed, 45 insertions(+), 179 deletions(-)
arch/x86/Kconfig +1 -0
@@ -115,6 +115,7 @@ config X86 select GENERIC_CPU_AUTOPROBE select GENERIC_CPU_VULNERABILITIES select GENERIC_EARLY_IOREMAP + select GENERIC_ENTRY select GENERIC_FIND_FIRST_BIT select GENERIC_IOMAP select GENERIC_IRQ_EFFECTIVE_AFF_MASK if SMP
arch/x86/entry/common.c +8 -173
252 lines not on this page · raw message
arch/x86/include/asm/entry-common.h new file +32 -0
@@ -0,0 +1,32 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +#ifndef _ASM_X86_ENTRY_COMMON_H +#define _ASM_X86_ENTRY_COMMON_H + +/* Check that the stack and regs on entry from user mode are sane. */ +static __always_inline void arch_check_user_regs(struct pt_regs *regs) +{ +if (IS_ENABLED (CONFIG_DEBUG_ENTRY )) { +/* +* Make sure that the entry code gave us a sensible EFLAGS +* register. Native because we want to check the actual CPU +* state, not the interrupt state as imagined by Xen. +*/ +unsigned long flags =native_save_fl (); +WARN_ON_ONCE (flags & (X86_EFLAGS_AC |X86_EFLAGS_DF | +X86_EFLAGS_NT )); + +/* We think we came from user mode. Make sure pt_regs agrees. */ +WARN_ON_ONCE (!user_mode (regs)); + +/* +* All entries from user mode (except #DF) should be on the +* normal thread stack and should have user pt_regs in the +* correct location. +*/ +WARN_ON_ONCE (!on_thread_stack ()); +WARN_ON_ONCE (regs !=task_pt_regs (current)); + } +} +#define arch_check_user_regs arch_check_user_regs + +#endif
arch/x86/include/asm/idtentry.h +4 -1
@@ -6,11 +6,14 @@#include <asm/trapnr.h> #ifndef __ASSEMBLY__ +#include <linux/entry-common.h> #include <linux/hardirq.h> #include <asm/irq_stack.h> -void idtentry_enter_user(struct pt_regs *regs); +/* Temporary define */ +#define idtentry_enter_user irqentry_enter_from_user_mode + void idtentry_exit_user(struct pt_regs *regs); typedef struct idtentry_state {
arch/x86/include/asm/thread_info.h +0 -5
@@ -133,11 +133,6 @@ struct thread_info {#define _TIF_X32 (1 << TIF_X32)#define _TIF_FSCHECK (1 << TIF_FSCHECK) -/* Work to do before invoking the actual syscall. */ -#define _TIF_WORK_SYSCALL_ENTRY \ - (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_EMU | _TIF_SYSCALL_AUDIT | \ - _TIF_SECCOMP | _TIF_SYSCALL_TRACEPOINT) -/* flags to check in __switch_to() */ #define _TIF_WORK_CTXSW_BASE \ (_TIF_NOCPUID | _TIF_NOTSC | _TIF_BLOCKSTEP | \
[tip: x86/entry] x86/entry: Use generic syscall entry function
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 27d6b4d14f5c3ab21c4aef87dd04055a2d7adf14 Gitweb: https://git.kernel.org/tip/27d6b4d14f5c3ab21c4aef87dd04055a2d7adf14 Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:04 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:59 +02:00
x86/entry: Use generic syscall entry function
Replace the syscall entry work handling with the generic version. Provide the necessary helper inlines to handle the real architecture specific parts, e.g. ptrace.
Use a temporary define for idtentry_enter_user which will be cleaned up seperately.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/Kconfig | 1 +- arch/x86/entry/common.c | 181 +-------------------------- arch/x86/include/asm/entry-common.h | 32 +++++- arch/x86/include/asm/idtentry.h | 5 +- arch/x86/include/asm/thread_info.h | 5 +- 5 files changed, 45 insertions(+), 179 deletions(-) create mode 100644 arch/x86/include/asm/entry-common.h
arch/x86/Kconfig +1 -0
@@ -115,6 +115,7 @@ config X86 select GENERIC_CPU_AUTOPROBE select GENERIC_CPU_VULNERABILITIES select GENERIC_EARLY_IOREMAP + select GENERIC_ENTRY select GENERIC_FIND_FIRST_BIT select GENERIC_IOMAP select GENERIC_IRQ_EFFECTIVE_AFF_MASK if SMP
arch/x86/entry/common.c +8 -173
252 lines not on this page · raw message
arch/x86/include/asm/entry-common.h new file +32 -0
@@ -0,0 +1,32 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +#ifndef _ASM_X86_ENTRY_COMMON_H +#define _ASM_X86_ENTRY_COMMON_H + +/* Check that the stack and regs on entry from user mode are sane. */ +static __always_inline void arch_check_user_regs(struct pt_regs *regs) +{ +if (IS_ENABLED (CONFIG_DEBUG_ENTRY )) { +/* +* Make sure that the entry code gave us a sensible EFLAGS +* register. Native because we want to check the actual CPU +* state, not the interrupt state as imagined by Xen. +*/ +unsigned long flags =native_save_fl (); +WARN_ON_ONCE (flags & (X86_EFLAGS_AC |X86_EFLAGS_DF | +X86_EFLAGS_NT )); + +/* We think we came from user mode. Make sure pt_regs agrees. */ +WARN_ON_ONCE (!user_mode (regs)); + +/* +* All entries from user mode (except #DF) should be on the +* normal thread stack and should have user pt_regs in the +* correct location. +*/ +WARN_ON_ONCE (!on_thread_stack ()); +WARN_ON_ONCE (regs !=task_pt_regs (current)); + } +} +#define arch_check_user_regs arch_check_user_regs + +#endif
arch/x86/include/asm/idtentry.h +4 -1
@@ -6,11 +6,14 @@#include <asm/trapnr.h> #ifndef __ASSEMBLY__ +#include <linux/entry-common.h> #include <linux/hardirq.h> #include <asm/irq_stack.h> -void idtentry_enter_user(struct pt_regs *regs); +/* Temporary define */ +#define idtentry_enter_user irqentry_enter_from_user_mode + void idtentry_exit_user(struct pt_regs *regs); typedef struct idtentry_state {
arch/x86/include/asm/thread_info.h +0 -5
@@ -133,11 +133,6 @@ struct thread_info {#define _TIF_X32 (1 << TIF_X32)#define _TIF_FSCHECK (1 << TIF_FSCHECK) -/* Work to do before invoking the actual syscall. */ -#define _TIF_WORK_SYSCALL_ENTRY \ - (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_EMU | _TIF_SYSCALL_AUDIT | \ - _TIF_SECCOMP | _TIF_SYSCALL_TRACEPOINT) -/* flags to check in __switch_to() */ #define _TIF_WORK_CTXSW_BASE \ (_TIF_NOCPUID | _TIF_NOTSC | _TIF_BLOCKSTEP | \
PATCH v5 11/15 x86/entry: Use generic syscall exit functionality
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Replace the x86 variant with the generic version. Provide the relevant architecture specific helper functions and defines.
Use a temporary define for idtentry_exit_user which will be cleaned up seperately.
Signed-off-by Thomas Gleixner <[email protected]>
Acked-by Kees Cook <[email protected]>
---
V4: Drop a pointless define
Adjust to moved TIF_USER_RETURN_NOTIFY handling Bring back I/O bitmap handling
--- arch/x86/entry/common.c | 221 ------------------------------------ arch/x86/entry/entry_32.S | 2 arch/x86/entry/entry_64.S | 2 arch/x86/include/asm/entry-common.h | 44 +++++++ arch/x86/include/asm/idtentry.h | 3 arch/x86/include/asm/signal.h | 1 arch/x86/kernel/signal.c | 3 7 files changed, 54 insertions(+), 222 deletions(-)
arch/x86/entry/common.c +5 -216
@@ -15,15 +15,8 @@#include <linux/smp.h> #include <linux/errno.h> #include <linux/ptrace.h> -#include <linux/tracehook.h> -#include <linux/audit.h> -#include <linux/signal.h> #include <linux/export.h> -#include <linux/context_tracking.h> -#include <linux/user-return-notifier.h> #include <linux/nospec.h> -#include <linux/uprobes.h> -#include <linux/livepatch.h> #include <linux/syscalls.h> #include <linux/uaccess.h>
@@ -42,191 +35,6 @@#include <asm/syscall.h> #include <asm/irq_stack.h> -#include <trace/events/syscalls.h> - -/** -* exit_to_user_mode - Fixup state when exiting to user mode -* -* Syscall exit enables interrupts, but the kernel state is interrupts -* disabled when this is invoked. Also tell RCU about it. -* -* 1) Trace interrupts on state -* 2) Invoke context tracking if enabled to adjust RCU state -* 3) Clear CPU buffers if CPU is affected by MDS and the migitation is on. -* 4) Tell lockdep that interrupts are enabled -*/ -static __always_inline void exit_to_user_mode(void) -{ -instrumentation_begin (); -trace_hardirqs_on_prepare (); -lockdep_hardirqs_on_prepare (CALLER_ADDR0 ); -instrumentation_end (); - -user_enter_irqoff (); -mds_user_clear_cpu_buffers (); -lockdep_hardirqs_on (CALLER_ADDR0 ); -} - -#define EXIT_TO_USERMODE_LOOP_FLAGS \ - (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \ - _TIF_NEED_RESCHED | _TIF_PATCH_PENDING) - -static void exit_to_usermode_loop (struct pt_regs *regs,u32 cached_flags) -{ -/* -* In order to return to user mode, we need to have IRQs off with -* none of EXIT_TO_USERMODE_LOOP_FLAGS set. Several of these flags -* can be set at any time on preemptible kernels if we have IRQs on, -* so we need to loop. Disabling preemption wouldn't help: doing the -* work to clear some of the flags can sleep. -*/ -while (true) { -/* We have work to do. */ -local_irq_enable (); - -if (cached_flags & _TIF_NEED_RESCHED) -schedule (); - -if (cached_flags & _TIF_UPROBE) -uprobe_notify_resume (regs); - -if (cached_flags & _TIF_PATCH_PENDING) -klp_update_patch_state (current); - -/* deal with pending signal delivery */ -if (cached_flags & _TIF_SIGPENDING) -do_signal (regs); - -if (cached_flags & _TIF_NOTIFY_RESUME) { -clear_thread_flag (TIF_NOTIFY_RESUME ); -tracehook_notify_resume (regs); -rseq_handle_notify_resume (NULL , regs); - } - -/* Disable IRQs and retry */ -local_irq_disable (); - - cached_flags =READ_ONCE (current_thread_info ()->flags); - -if (!(cached_flags &EXIT_TO_USERMODE_LOOP_FLAGS )) -break ; - } -} - -static void __prepare_exit_to_usermode (struct pt_regs *regs) -{ -struct thread_info *ti =current_thread_info (); -u32 cached_flags; - -addr_limit_user_check (); - -lockdep_assert_irqs_disabled (); -lockdep_sys_exit (); - - cached_flags =READ_ONCE (ti->flags); - -if (unlikely (cached_flags &EXIT_TO_USERMODE_LOOP_FLAGS )) -exit_to_usermode_loop (regs, cached_flags); - -/* Reload ti->flags; we may have rescheduled above. */ - cached_flags =READ_ONCE (ti->flags); - -if (cached_flags & _TIF_USER_RETURN_NOTIFY) -fire_user_return_notifiers (); - -if (unlikely (cached_flags & _TIF_IO_BITMAP)) -tss_update_io_bitmap (); - -fpregs_assert_state_consistent (); -if (unlikely (cached_flags & _TIF_NEED_FPU_LOAD)) -switch_fpu_return (); - -#ifdef CONFIG_COMPAT -/* -* Compat syscalls set TS_COMPAT. Make sure we clear it before -* returning to user mode. We need to clear it *after* signal -* handling, because syscall restart has a fixup for compat -* syscalls. The fixup is exercised by the ptrace_syscall_32 -* selftest. -* -* We also need to clear TS_REGS_POKED_I386: the 32-bit tracer -* special case only applies after poking regs and before the -* very next return to user mode. -*/ - ti->status &= ~(TS_COMPAT |TS_I386_REGS_POKED ); -#endif -} - -static noinstr void prepare_exit_to_usermode(struct pt_regs *regs) -{ -instrumentation_begin (); -__prepare_exit_to_usermode (regs); -instrumentation_end (); -exit_to_user_mode (); -} - -#define SYSCALL_EXIT_WORK_FLAGS \ - (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | \ - _TIF_SINGLESTEP | _TIF_SYSCALL_TRACEPOINT) - -static void syscall_slow_exit_work (struct pt_regs *regs,u32 cached_flags) -{ -bool step; - -audit_syscall_exit (regs); - -if (cached_flags & _TIF_SYSCALL_TRACEPOINT) -trace_sys_exit (regs, regs->ax); - -/* -* If TIF_SYSCALL_EMU is set, we only get here because of -* TIF_SINGLESTEP (i.e. this is PTRACE_SYSEMU_SINGLESTEP). -* We already reported this syscall instruction in -* syscall_trace_enter(). -*/ - step =unlikely ( - (cached_flags & (_TIF_SINGLESTEP | _TIF_SYSCALL_EMU)) - == _TIF_SINGLESTEP); -if (step || cached_flags & _TIF_SYSCALL_TRACE) -tracehook_report_syscall_exit (regs, step); -} - -static void __syscall_return_slowpath (struct pt_regs *regs) -{ -struct thread_info *ti =current_thread_info (); -u32 cached_flags =READ_ONCE (ti->flags); - -CT_WARN_ON (ct_state () !=CONTEXT_KERNEL ); - -if (IS_ENABLED (CONFIG_PROVE_LOCKING ) && -WARN (irqs_disabled (),"syscall %ld left IRQs disabled" , regs->orig_ax)) -local_irq_enable (); - -rseq_syscall (regs); - -/* -* First do one-time work. If these work items are enabled, we -* want to run them exactly once per syscall exit with IRQs on. -*/ -if (unlikely (cached_flags &SYSCALL_EXIT_WORK_FLAGS )) -syscall_slow_exit_work (regs, cached_flags); - -local_irq_disable (); -__prepare_exit_to_usermode (regs); -} - -/* -* Called with IRQs on and fully valid regs. Returns with IRQs off in a -* state such that we can immediately switch to user mode. -*/ -__visible noinstr void syscall_return_slowpath(struct pt_regs *regs) -{ -instrumentation_begin (); -__syscall_return_slowpath (regs); -instrumentation_end (); -exit_to_user_mode (); -} -#ifdef CONFIG_X86_64 __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) {
@@ -245,7 +53,7 @@ static void __syscall_return_slowpath(st #endif }instrumentation_end (); -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs); }#endif
@@ -284,7 +92,7 @@ static __always_inline void do_syscall_3unsigned int nr =syscall_32_enter (regs);do_syscall_32_irqs_on (regs, nr); -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs); } static noinstr bool __do_fast_syscall_32(struct pt_regs *regs)
@@ -310,13 +118,13 @@ static noinstr bool __do_fast_syscall_32if (res) {/* User code screwed up. */ regs->ax = -EFAULT ; -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs);return false; }/* Now this is just like a normal syscall. */ do_syscall_32_irqs_on (regs, nr); -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs);return true; }
@@ -524,7 +332,7 @@ void noinstr idtentry_exit(struct pt_reg/* Check whether this returns to user mode */ if (user_mode (regs)) { -prepare_exit_to_usermode (regs); +irqentry_exit_to_user_mode (regs); }else if (regs->flags &X86_EFLAGS_IF ) {/* * If RCU was not watching on entry this needs to be done
@@ -555,25 +363,6 @@ void noinstr idtentry_exit(struct pt_reg } } -/** -* idtentry_exit_user - Handle return from exception to user mode -* @regs: Pointer to pt_regs (exception entry regs) -* -* Runs the necessary preemption and work checks and returns to the caller -* with interrupts disabled and no further work pending. -* -* This is the last action before returning to the low level ASM code which -* just needs to return to the appropriate context. -* -* Counterpart to idtentry_enter_user(). -*/ -void noinstr idtentry_exit_user(struct pt_regs *regs) -{ -lockdep_assert_irqs_disabled (); - -prepare_exit_to_usermode (regs); -} -#ifdef CONFIG_XEN_PV #ifndef CONFIG_PREEMPTION /*
arch/x86/entry/entry_32.S +1 -1
@@ -846,7 +846,7 @@ SYM_CODE_START(ret_from_fork) 2: /* When we fork, we trace the syscall return in the child, too. */ movl %esp, %eax - call syscall_return_slowpath + call syscall_exit_to_user_mode jmp .Lsyscall_32_done /* kernel thread */
arch/x86/entry/entry_64.S +1 -1
@@ -283,7 +283,7 @@ SYM_CODE_START(ret_from_fork) 2: UNWIND_HINT_REGS movq %rsp, %rdi - call syscall_return_slowpath /* returns with IRQs disabled */ + call syscall_exit_to_user_mode /* returns with IRQs disabled */ jmp swapgs_restore_regs_and_return_to_usermode 1:
arch/x86/include/asm/entry-common.h +44 -0
54 lines not on this page · raw message
arch/x86/include/asm/idtentry.h +1 -2
@@ -13,8 +13,7 @@/* Temporary define */ #define idtentry_enter_user irqentry_enter_from_user_mode - -void idtentry_exit_user (struct pt_regs *regs); +#define idtentry_exit_user irqentry_exit_to_user_mode typedef struct idtentry_state { bool exit_rcu;
arch/x86/include/asm/signal.h +0 -1
@@ -35,7 +35,6 @@ typedef sigset_t compat_sigset_t;#endif /* __ASSEMBLY__ */ #include <uapi/asm/signal.h> #ifndef __ASSEMBLY__ -extern void do_signal (struct pt_regs *regs);#define __ARCH_HAS_SA_RESTORER
arch/x86/kernel/signal.c +2 -1
@@ -25,6 +25,7 @@#include <linux/user-return-notifier.h> #include <linux/uprobes.h> #include <linux/context_tracking.h> +#include <linux/entry-common.h> #include <linux/syscalls.h> #include <asm/processor.h>
@@ -803,7 +804,7 @@ static inline unsigned long get_nr_resta* want to handle. Thus you cannot kill init even with a SIGKILL even by * mistake. */ -void do_signal (struct pt_regs *regs) +void arch_do_signal (struct pt_regs *regs) {struct ksignal ksig;
[tip: x86/entry] x86/entry: Use generic syscall exit functionality
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 167fd210ec0555d371a20435dac7c2c7052df7ed Gitweb: https://git.kernel.org/tip/167fd210ec0555d371a20435dac7c2c7052df7ed Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:05 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:04:59 +02:00
x86/entry: Use generic syscall exit functionality
Replace the x86 variant with the generic version. Provide the relevant architecture specific helper functions and defines.
Use a temporary define for idtentry_exit_user which will be cleaned up seperately.
Signed-off-by Thomas Gleixner <[email protected]>
Acked-by Kees Cook <[email protected]>
--- arch/x86/entry/common.c | 221 +--------------------------- arch/x86/entry/entry_32.S | 2 +- arch/x86/entry/entry_64.S | 2 +- arch/x86/include/asm/entry-common.h | 44 +++++- arch/x86/include/asm/idtentry.h | 3 +- arch/x86/include/asm/signal.h | 1 +- arch/x86/kernel/signal.c | 3 +- 7 files changed, 54 insertions(+), 222 deletions(-)
arch/x86/entry/common.c +5 -216
@@ -15,15 +15,8 @@#include <linux/smp.h> #include <linux/errno.h> #include <linux/ptrace.h> -#include <linux/tracehook.h> -#include <linux/audit.h> -#include <linux/signal.h> #include <linux/export.h> -#include <linux/context_tracking.h> -#include <linux/user-return-notifier.h> #include <linux/nospec.h> -#include <linux/uprobes.h> -#include <linux/livepatch.h> #include <linux/syscalls.h> #include <linux/uaccess.h>
@@ -42,191 +35,6 @@#include <asm/syscall.h> #include <asm/irq_stack.h> -#include <trace/events/syscalls.h> - -/** -* exit_to_user_mode - Fixup state when exiting to user mode -* -* Syscall exit enables interrupts, but the kernel state is interrupts -* disabled when this is invoked. Also tell RCU about it. -* -* 1) Trace interrupts on state -* 2) Invoke context tracking if enabled to adjust RCU state -* 3) Clear CPU buffers if CPU is affected by MDS and the migitation is on. -* 4) Tell lockdep that interrupts are enabled -*/ -static __always_inline void exit_to_user_mode(void) -{ -instrumentation_begin (); -trace_hardirqs_on_prepare (); -lockdep_hardirqs_on_prepare (CALLER_ADDR0 ); -instrumentation_end (); - -user_enter_irqoff (); -mds_user_clear_cpu_buffers (); -lockdep_hardirqs_on (CALLER_ADDR0 ); -} - -#define EXIT_TO_USERMODE_LOOP_FLAGS \ - (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \ - _TIF_NEED_RESCHED | _TIF_PATCH_PENDING) - -static void exit_to_usermode_loop (struct pt_regs *regs,u32 cached_flags) -{ -/* -* In order to return to user mode, we need to have IRQs off with -* none of EXIT_TO_USERMODE_LOOP_FLAGS set. Several of these flags -* can be set at any time on preemptible kernels if we have IRQs on, -* so we need to loop. Disabling preemption wouldn't help: doing the -* work to clear some of the flags can sleep. -*/ -while (true) { -/* We have work to do. */ -local_irq_enable (); - -if (cached_flags & _TIF_NEED_RESCHED) -schedule (); - -if (cached_flags & _TIF_UPROBE) -uprobe_notify_resume (regs); - -if (cached_flags & _TIF_PATCH_PENDING) -klp_update_patch_state (current); - -/* deal with pending signal delivery */ -if (cached_flags & _TIF_SIGPENDING) -do_signal (regs); - -if (cached_flags & _TIF_NOTIFY_RESUME) { -clear_thread_flag (TIF_NOTIFY_RESUME ); -tracehook_notify_resume (regs); -rseq_handle_notify_resume (NULL , regs); - } - -/* Disable IRQs and retry */ -local_irq_disable (); - - cached_flags =READ_ONCE (current_thread_info ()->flags); - -if (!(cached_flags &EXIT_TO_USERMODE_LOOP_FLAGS )) -break ; - } -} - -static void __prepare_exit_to_usermode (struct pt_regs *regs) -{ -struct thread_info *ti =current_thread_info (); -u32 cached_flags; - -addr_limit_user_check (); - -lockdep_assert_irqs_disabled (); -lockdep_sys_exit (); - - cached_flags =READ_ONCE (ti->flags); - -if (unlikely (cached_flags &EXIT_TO_USERMODE_LOOP_FLAGS )) -exit_to_usermode_loop (regs, cached_flags); - -/* Reload ti->flags; we may have rescheduled above. */ - cached_flags =READ_ONCE (ti->flags); - -if (cached_flags & _TIF_USER_RETURN_NOTIFY) -fire_user_return_notifiers (); - -if (unlikely (cached_flags & _TIF_IO_BITMAP)) -tss_update_io_bitmap (); - -fpregs_assert_state_consistent (); -if (unlikely (cached_flags & _TIF_NEED_FPU_LOAD)) -switch_fpu_return (); - -#ifdef CONFIG_COMPAT -/* -* Compat syscalls set TS_COMPAT. Make sure we clear it before -* returning to user mode. We need to clear it *after* signal -* handling, because syscall restart has a fixup for compat -* syscalls. The fixup is exercised by the ptrace_syscall_32 -* selftest. -* -* We also need to clear TS_REGS_POKED_I386: the 32-bit tracer -* special case only applies after poking regs and before the -* very next return to user mode. -*/ - ti->status &= ~(TS_COMPAT |TS_I386_REGS_POKED ); -#endif -} - -static noinstr void prepare_exit_to_usermode(struct pt_regs *regs) -{ -instrumentation_begin (); -__prepare_exit_to_usermode (regs); -instrumentation_end (); -exit_to_user_mode (); -} - -#define SYSCALL_EXIT_WORK_FLAGS \ - (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | \ - _TIF_SINGLESTEP | _TIF_SYSCALL_TRACEPOINT) - -static void syscall_slow_exit_work (struct pt_regs *regs,u32 cached_flags) -{ -bool step; - -audit_syscall_exit (regs); - -if (cached_flags & _TIF_SYSCALL_TRACEPOINT) -trace_sys_exit (regs, regs->ax); - -/* -* If TIF_SYSCALL_EMU is set, we only get here because of -* TIF_SINGLESTEP (i.e. this is PTRACE_SYSEMU_SINGLESTEP). -* We already reported this syscall instruction in -* syscall_trace_enter(). -*/ - step =unlikely ( - (cached_flags & (_TIF_SINGLESTEP | _TIF_SYSCALL_EMU)) - == _TIF_SINGLESTEP); -if (step || cached_flags & _TIF_SYSCALL_TRACE) -tracehook_report_syscall_exit (regs, step); -} - -static void __syscall_return_slowpath (struct pt_regs *regs) -{ -struct thread_info *ti =current_thread_info (); -u32 cached_flags =READ_ONCE (ti->flags); - -CT_WARN_ON (ct_state () !=CONTEXT_KERNEL ); - -if (IS_ENABLED (CONFIG_PROVE_LOCKING ) && -WARN (irqs_disabled (),"syscall %ld left IRQs disabled" , regs->orig_ax)) -local_irq_enable (); - -rseq_syscall (regs); - -/* -* First do one-time work. If these work items are enabled, we -* want to run them exactly once per syscall exit with IRQs on. -*/ -if (unlikely (cached_flags &SYSCALL_EXIT_WORK_FLAGS )) -syscall_slow_exit_work (regs, cached_flags); - -local_irq_disable (); -__prepare_exit_to_usermode (regs); -} - -/* -* Called with IRQs on and fully valid regs. Returns with IRQs off in a -* state such that we can immediately switch to user mode. -*/ -__visible noinstr void syscall_return_slowpath(struct pt_regs *regs) -{ -instrumentation_begin (); -__syscall_return_slowpath (regs); -instrumentation_end (); -exit_to_user_mode (); -} -#ifdef CONFIG_X86_64 __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) {
@@ -245,7 +53,7 @@ __visible noinstr void do_syscall_64(unsigned long nr, struct pt_regs *regs) #endif }instrumentation_end (); -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs); }#endif
@@ -284,7 +92,7 @@ __visible noinstr void do_int80_syscall_32(struct pt_regs *regs)unsigned int nr =syscall_32_enter (regs);do_syscall_32_irqs_on (regs, nr); -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs); } static noinstr bool __do_fast_syscall_32(struct pt_regs *regs)
@@ -310,13 +118,13 @@ static noinstr bool __do_fast_syscall_32(struct pt_regs *regs)if (res) {/* User code screwed up. */ regs->ax = -EFAULT ; -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs);return false; }/* Now this is just like a normal syscall. */ do_syscall_32_irqs_on (regs, nr); -syscall_return_slowpath (regs); +syscall_exit_to_user_mode (regs);return true; }
@@ -524,7 +332,7 @@ void noinstr idtentry_exit(struct pt_regs *regs, idtentry_state_t state)/* Check whether this returns to user mode */ if (user_mode (regs)) { -prepare_exit_to_usermode (regs); +irqentry_exit_to_user_mode (regs); }else if (regs->flags &X86_EFLAGS_IF ) {/* * If RCU was not watching on entry this needs to be done
@@ -555,25 +363,6 @@ void noinstr idtentry_exit(struct pt_regs *regs, idtentry_state_t state) } } -/** -* idtentry_exit_user - Handle return from exception to user mode -* @regs: Pointer to pt_regs (exception entry regs) -* -* Runs the necessary preemption and work checks and returns to the caller -* with interrupts disabled and no further work pending. -* -* This is the last action before returning to the low level ASM code which -* just needs to return to the appropriate context. -* -* Counterpart to idtentry_enter_user(). -*/ -void noinstr idtentry_exit_user(struct pt_regs *regs) -{ -lockdep_assert_irqs_disabled (); - -prepare_exit_to_usermode (regs); -} -#ifdef CONFIG_XEN_PV #ifndef CONFIG_PREEMPTION /*
arch/x86/entry/entry_32.S +1 -1
@@ -846,7 +846,7 @@ SYM_CODE_START(ret_from_fork) 2: /* When we fork, we trace the syscall return in the child, too. */ movl %esp, %eax - call syscall_return_slowpath + call syscall_exit_to_user_mode jmp .Lsyscall_32_done /* kernel thread */
arch/x86/entry/entry_64.S +1 -1
@@ -283,7 +283,7 @@ SYM_CODE_START(ret_from_fork) 2: UNWIND_HINT_REGS movq %rsp, %rdi - call syscall_return_slowpath /* returns with IRQs disabled */ + call syscall_exit_to_user_mode /* returns with IRQs disabled */ jmp swapgs_restore_regs_and_return_to_usermode 1:
arch/x86/include/asm/entry-common.h +44 -0
54 lines not on this page · raw message
arch/x86/include/asm/idtentry.h +1 -2
@@ -13,8 +13,7 @@/* Temporary define */ #define idtentry_enter_user irqentry_enter_from_user_mode - -void idtentry_exit_user (struct pt_regs *regs); +#define idtentry_exit_user irqentry_exit_to_user_mode typedef struct idtentry_state { bool exit_rcu;
arch/x86/include/asm/signal.h +0 -1
@@ -35,7 +35,6 @@ typedef sigset_t compat_sigset_t;#endif /* __ASSEMBLY__ */ #include <uapi/asm/signal.h> #ifndef __ASSEMBLY__ -extern void do_signal (struct pt_regs *regs);#define __ARCH_HAS_SA_RESTORER
arch/x86/kernel/signal.c +2 -1
@@ -25,6 +25,7 @@#include <linux/user-return-notifier.h> #include <linux/uprobes.h> #include <linux/context_tracking.h> +#include <linux/entry-common.h> #include <linux/syscalls.h> #include <asm/processor.h>
@@ -803,7 +804,7 @@ static inline unsigned long get_nr_restart_syscall(const struct pt_regs *regs)* want to handle. Thus you cannot kill init even with a SIGKILL even by * mistake. */ -void do_signal (struct pt_regs *regs) +void arch_do_signal (struct pt_regs *regs) {struct ksignal ksig;
PATCH v5 12/15 x86/entry: Cleanup idtentry_entry/exit_user
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Cleanup the temporary defines and use irqentry_ instead of idtentry_.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/include/asm/idtentry.h | 4 ---- arch/x86/kernel/cpu/mce/core.c | 4 ++-- arch/x86/kernel/traps.c | 18 +++++++++--------- 3 files changed, 11 insertions(+), 15 deletions(-)
arch/x86/include/asm/idtentry.h +0 -4
@@ -11,10 +11,6 @@#include <asm/irq_stack.h> -/* Temporary define */ -#define idtentry_enter_user irqentry_enter_from_user_mode -#define idtentry_exit_user irqentry_exit_to_user_mode -typedef struct idtentry_state {bool exit_rcu; }idtentry_state_t ;
arch/x86/kernel/cpu/mce/core.c +2 -2
@@ -1927,11 +1927,11 @@ static __always_inline void exc_machine_ static __always_inline void exc_machine_check_user(struct pt_regs *regs) { -idtentry_enter_user (regs); +irqentry_enter_from_user_mode (regs);instrumentation_begin ();machine_check_vector (regs);instrumentation_end (); -idtentry_exit_user (regs); +irqentry_exit_to_user_mode (regs); }#ifdef CONFIG_X86_64
arch/x86/kernel/traps.c +9 -9
@@ -638,18 +638,18 @@ DEFINE_IDTENTRY_RAW(exc_int3)return ;/* -* idtentry_enter_user() uses static_branch_{,un}likely() and therefore -* can trigger INT3, hence poke_int3_handler() must be done -* before. If the entry came from kernel mode, then use nmi_enter() -* because the INT3 could have been hit in any context including -* NMI. +* irqentry_enter_from_user_mode() uses static_branch_{,un}likely() +* and therefore can trigger INT3, hence poke_int3_handler() must +* be done before. If the entry came from kernel mode, then use +* nmi_enter() because the INT3 could have been hit in any context +* including NMI. */ if (user_mode (regs)) { -idtentry_enter_user (regs); +irqentry_enter_from_user_mode (regs);instrumentation_begin ();do_int3_user (regs);instrumentation_end (); -idtentry_exit_user (regs); +irqentry_exit_to_user_mode (regs); }else {nmi_enter ();instrumentation_begin ();
@@ -901,12 +901,12 @@ static __always_inline void exc_debug_us*/ WARN_ON_ONCE (!user_mode (regs)); -idtentry_enter_user (regs); +irqentry_enter_from_user_mode (regs);instrumentation_begin ();handle_debug (regs, dr6, true);instrumentation_end (); -idtentry_exit_user (regs); +irqentry_exit_to_user_mode (regs); }#ifdef CONFIG_X86_64
[tip: x86/entry] x86/entry: Cleanup idtentry_entry/exit_user
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 517e499227bed34acc69166691e2db5df3dc859a Gitweb: https://git.kernel.org/tip/517e499227bed34acc69166691e2db5df3dc859a Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:06 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:05:00 +02:00
x86/entry: Cleanup idtentry_entry/exit_user
Cleanup the temporary defines and use irqentry_ instead of idtentry_.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/include/asm/idtentry.h | 4 ---- arch/x86/kernel/cpu/mce/core.c | 4 ++-- arch/x86/kernel/traps.c | 18 +++++++++--------- 3 files changed, 11 insertions(+), 15 deletions(-)
arch/x86/include/asm/idtentry.h +0 -4
@@ -11,10 +11,6 @@#include <asm/irq_stack.h> -/* Temporary define */ -#define idtentry_enter_user irqentry_enter_from_user_mode -#define idtentry_exit_user irqentry_exit_to_user_mode -typedef struct idtentry_state {bool exit_rcu; }idtentry_state_t ;
arch/x86/kernel/cpu/mce/core.c +2 -2
@@ -1927,11 +1927,11 @@ static __always_inline void exc_machine_check_kernel(struct pt_regs *regs) static __always_inline void exc_machine_check_user(struct pt_regs *regs) { -idtentry_enter_user (regs); +irqentry_enter_from_user_mode (regs);instrumentation_begin ();machine_check_vector (regs);instrumentation_end (); -idtentry_exit_user (regs); +irqentry_exit_to_user_mode (regs); }#ifdef CONFIG_X86_64
arch/x86/kernel/traps.c +9 -9
@@ -638,18 +638,18 @@ DEFINE_IDTENTRY_RAW(exc_int3)return ;/* -* idtentry_enter_user() uses static_branch_{,un}likely() and therefore -* can trigger INT3, hence poke_int3_handler() must be done -* before. If the entry came from kernel mode, then use nmi_enter() -* because the INT3 could have been hit in any context including -* NMI. +* irqentry_enter_from_user_mode() uses static_branch_{,un}likely() +* and therefore can trigger INT3, hence poke_int3_handler() must +* be done before. If the entry came from kernel mode, then use +* nmi_enter() because the INT3 could have been hit in any context +* including NMI. */ if (user_mode (regs)) { -idtentry_enter_user (regs); +irqentry_enter_from_user_mode (regs);instrumentation_begin ();do_int3_user (regs);instrumentation_end (); -idtentry_exit_user (regs); +irqentry_exit_to_user_mode (regs); }else {nmi_enter ();instrumentation_begin ();
@@ -901,12 +901,12 @@ static __always_inline void exc_debug_user(struct pt_regs *regs,*/ WARN_ON_ONCE (!user_mode (regs)); -idtentry_enter_user (regs); +irqentry_enter_from_user_mode (regs);instrumentation_begin ();handle_debug (regs, dr6, true);instrumentation_end (); -idtentry_exit_user (regs); +irqentry_exit_to_user_mode (regs); }#ifdef CONFIG_X86_64
PATCH v5 13/15 x86/entry: Use generic interrupt entry/exit code
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Replace the x86 code with the generic variant. Use temporary defines for idtentry_* which will be cleaned up in the next step.
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/entry/common.c | 167 ---------------------------------------- arch/x86/include/asm/idtentry.h | 10 -- 2 files changed, 5 insertions(+), 172 deletions(-)
arch/x86/entry/common.c +1 -166
@@ -198,171 +198,6 @@ SYSCALL_DEFINE0(ni_syscall)return -ENOSYS ; } -/** -* idtentry_enter - Handle state tracking on ordinary idtentries -* @regs: Pointer to pt_regs of interrupted context -* -* Invokes: -* - lockdep irqflag state tracking as low level ASM entry disabled -* interrupts. -* -* - Context tracking if the exception hit user mode. -* -* - The hardirq tracer to keep the state consistent as low level ASM -* entry disabled interrupts. -* -* As a precondition, this requires that the entry came from user mode, -* idle, or a kernel context in which RCU is watching. -* -* For kernel mode entries RCU handling is done conditional. If RCU is -* watching then the only RCU requirement is to check whether the tick has -* to be restarted. If RCU is not watching then rcu_irq_enter() has to be -* invoked on entry and rcu_irq_exit() on exit. -* -* Avoiding the rcu_irq_enter/exit() calls is an optimization but also -* solves the problem of kernel mode pagefaults which can schedule, which -* is not possible after invoking rcu_irq_enter() without undoing it. -* -* For user mode entries irqentry_enter_from_user_mode() must be invoked to -* establish the proper context for NOHZ_FULL. Otherwise scheduling on exit -* would not be possible. -* -* Returns: An opaque object that must be passed to idtentry_exit() -* -* The return value must be fed into the state argument of -* idtentry_exit(). -*/ -idtentry_state_t noinstr idtentry_enter(struct pt_regs *regs) -{ -idtentry_state_t ret = { - .exit_rcu = false, - }; - -if (user_mode (regs)) { -irqentry_enter_from_user_mode (regs); -return ret; - } - -/* -* If this entry hit the idle task invoke rcu_irq_enter() whether -* RCU is watching or not. -* -* Interupts can nest when the first interrupt invokes softirq -* processing on return which enables interrupts. -* -* Scheduler ticks in the idle task can mark quiescent state and -* terminate a grace period, if and only if the timer interrupt is -* not nested into another interrupt. -* -* Checking for __rcu_is_watching() here would prevent the nesting -* interrupt to invoke rcu_irq_enter(). If that nested interrupt is -* the tick then rcu_flavor_sched_clock_irq() would wrongfully -* assume that it is the first interupt and eventually claim -* quiescient state and end grace periods prematurely. -* -* Unconditionally invoke rcu_irq_enter() so RCU state stays -* consistent. -* -* TINY_RCU does not support EQS, so let the compiler eliminate -* this part when enabled. -*/ -if (!IS_ENABLED (CONFIG_TINY_RCU ) &&is_idle_task (current)) { -/* -* If RCU is not watching then the same careful -* sequence vs. lockdep and tracing is required -* as in irqentry_enter_from_user_mode(). -*/ -lockdep_hardirqs_off (CALLER_ADDR0 ); -rcu_irq_enter (); -instrumentation_begin (); -trace_hardirqs_off_finish (); -instrumentation_end (); - - ret.exit_rcu = true; -return ret; - } - -/* -* If RCU is watching then RCU only wants to check whether it needs -* to restart the tick in NOHZ mode. rcu_irq_enter_check_tick() -* already contains a warning when RCU is not watching, so no point -* in having another one here. -*/ -instrumentation_begin (); -rcu_irq_enter_check_tick (); -/* Use the combo lockdep/tracing function */ -trace_hardirqs_off (); -instrumentation_end (); - -return ret; -} - -static void idtentry_exit_cond_resched (struct pt_regs *regs,bool may_sched) -{ -if (may_sched && !preempt_count ()) { -/* Sanity check RCU and thread stack */ -rcu_irq_exit_check_preempt (); -if (IS_ENABLED (CONFIG_DEBUG_ENTRY )) -WARN_ON_ONCE (!on_thread_stack ()); -if (need_resched ()) -preempt_schedule_irq (); - } -/* Covers both tracing and lockdep */ -trace_hardirqs_on (); -} - -/** -* idtentry_exit - Handle return from exception that used idtentry_enter() -* @regs: Pointer to pt_regs (exception entry regs) -* @state: Return value from matching call to idtentry_enter() -* -* Depending on the return target (kernel/user) this runs the necessary -* preemption and work checks if possible and reguired and returns to -* the caller with interrupts disabled and no further work pending. -* -* This is the last action before returning to the low level ASM code which -* just needs to return to the appropriate context. -* -* Counterpart to idtentry_enter(). The return value of the entry -* function must be fed into the @state argument. -*/ -void noinstr idtentry_exit(struct pt_regs *regs, idtentry_state_t state) -{ -lockdep_assert_irqs_disabled (); - -/* Check whether this returns to user mode */ -if (user_mode (regs)) { -irqentry_exit_to_user_mode (regs); - }else if (regs->flags &X86_EFLAGS_IF ) { -/* -* If RCU was not watching on entry this needs to be done -* carefully and needs the same ordering of lockdep/tracing -* and RCU as the return to user mode path. -*/ -if (state.exit_rcu) { -instrumentation_begin (); -/* Tell the tracer that IRET will enable interrupts */ -trace_hardirqs_on_prepare (); -lockdep_hardirqs_on_prepare (CALLER_ADDR0 ); -instrumentation_end (); -rcu_irq_exit (); -lockdep_hardirqs_on (CALLER_ADDR0 ); -return ; - } - -instrumentation_begin (); -idtentry_exit_cond_resched (regs,IS_ENABLED (CONFIG_PREEMPTION )); -instrumentation_end (); - }else { -/* -* IRQ flags state is correct already. Just tell RCU if it -* was not watching on entry. -*/ -if (state.exit_rcu) -rcu_irq_exit (); - } -} -#ifdef CONFIG_XEN_PV #ifndef CONFIG_PREEMPTION /*
@@ -427,7 +262,7 @@ static void __xen_pv_evtchn_do_upcall(vo inhcall =get_and_clear_inhcall ();if (inhcall && !WARN_ON_ONCE (state.exit_rcu)) {instrumentation_begin (); -idtentry_exit_cond_resched (regs, true); +irqentry_exit_cond_resched ();instrumentation_end ();restore_inhcall (inhcall); }else {
arch/x86/include/asm/idtentry.h +4 -6
@@ -11,12 +11,10 @@#include <asm/irq_stack.h> -typedef struct idtentry_state { -bool exit_rcu; -}idtentry_state_t ; - -idtentry_state_t idtentry_enter (struct pt_regs *regs); -void idtentry_exit (struct pt_regs *regs,idtentry_state_t state); +/* Temporary defines */ +typedef irqentry_state_t idtentry_state_t ; +#define idtentry_enter irqentry_enter +#define idtentry_exit irqentry_exit/** * DECLARE_IDTENTRY - Declare functions for simple IDT entry points
Re: [patch V5 13/15] x86/entry: Use generic interrupt entry/exit code
Quoted 6 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Replace the x86 code with the generic variant. Use temporary defines for idtentry_* which will be cleaned up in the next step.
Signed-off-by Thomas Gleixner <[email protected]>
There was a comment that still referenced the old x86-specific API names - the patch below updates them.
Thanks,
Ingo
====
arch/x86/include/asm/idtentry.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/include/asm/idtentry.h +1 -1
@@ -128,7 +128,7 @@ static __always_inline void __##func(struct pt_regs *regs, \* body with a pair of curly brackets. * * Contrary to DEFINE_IDTENTRY() this does not invoke the -* idtentry_enter/exit() helpers before and after the body invocation. This +* irqentry_enter/exit() helpers before and after the body invocation. This * needs to be done in the body itself if applicable. Use if extra work * is required before the enter/exit() helpers are invoked. */
[tip: x86/entry] x86/entry: Use generic interrupt entry/exit code
In reply to Thomas GleixnerThe following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: bdcd178ada90d2413bcc9df4211dcdd511a47586 Gitweb: https://git.kernel.org/tip/bdcd178ada90d2413bcc9df4211dcdd511a47586 Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:07 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:05:00 +02:00
x86/entry: Use generic interrupt entry/exit code
Replace the x86 code with the generic variant. Use temporary defines for idtentry_* which will be cleaned up in the next step.
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/entry/common.c | 167 +------------------------------- arch/x86/include/asm/idtentry.h | 10 +-- 2 files changed, 5 insertions(+), 172 deletions(-)
arch/x86/entry/common.c +1 -166
@@ -198,171 +198,6 @@ SYSCALL_DEFINE0(ni_syscall)return -ENOSYS ; } -/** -* idtentry_enter - Handle state tracking on ordinary idtentries -* @regs: Pointer to pt_regs of interrupted context -* -* Invokes: -* - lockdep irqflag state tracking as low level ASM entry disabled -* interrupts. -* -* - Context tracking if the exception hit user mode. -* -* - The hardirq tracer to keep the state consistent as low level ASM -* entry disabled interrupts. -* -* As a precondition, this requires that the entry came from user mode, -* idle, or a kernel context in which RCU is watching. -* -* For kernel mode entries RCU handling is done conditional. If RCU is -* watching then the only RCU requirement is to check whether the tick has -* to be restarted. If RCU is not watching then rcu_irq_enter() has to be -* invoked on entry and rcu_irq_exit() on exit. -* -* Avoiding the rcu_irq_enter/exit() calls is an optimization but also -* solves the problem of kernel mode pagefaults which can schedule, which -* is not possible after invoking rcu_irq_enter() without undoing it. -* -* For user mode entries irqentry_enter_from_user_mode() must be invoked to -* establish the proper context for NOHZ_FULL. Otherwise scheduling on exit -* would not be possible. -* -* Returns: An opaque object that must be passed to idtentry_exit() -* -* The return value must be fed into the state argument of -* idtentry_exit(). -*/ -idtentry_state_t noinstr idtentry_enter(struct pt_regs *regs) -{ -idtentry_state_t ret = { - .exit_rcu = false, - }; - -if (user_mode (regs)) { -irqentry_enter_from_user_mode (regs); -return ret; - } - -/* -* If this entry hit the idle task invoke rcu_irq_enter() whether -* RCU is watching or not. -* -* Interupts can nest when the first interrupt invokes softirq -* processing on return which enables interrupts. -* -* Scheduler ticks in the idle task can mark quiescent state and -* terminate a grace period, if and only if the timer interrupt is -* not nested into another interrupt. -* -* Checking for __rcu_is_watching() here would prevent the nesting -* interrupt to invoke rcu_irq_enter(). If that nested interrupt is -* the tick then rcu_flavor_sched_clock_irq() would wrongfully -* assume that it is the first interupt and eventually claim -* quiescient state and end grace periods prematurely. -* -* Unconditionally invoke rcu_irq_enter() so RCU state stays -* consistent. -* -* TINY_RCU does not support EQS, so let the compiler eliminate -* this part when enabled. -*/ -if (!IS_ENABLED (CONFIG_TINY_RCU ) &&is_idle_task (current)) { -/* -* If RCU is not watching then the same careful -* sequence vs. lockdep and tracing is required -* as in irqentry_enter_from_user_mode(). -*/ -lockdep_hardirqs_off (CALLER_ADDR0 ); -rcu_irq_enter (); -instrumentation_begin (); -trace_hardirqs_off_finish (); -instrumentation_end (); - - ret.exit_rcu = true; -return ret; - } - -/* -* If RCU is watching then RCU only wants to check whether it needs -* to restart the tick in NOHZ mode. rcu_irq_enter_check_tick() -* already contains a warning when RCU is not watching, so no point -* in having another one here. -*/ -instrumentation_begin (); -rcu_irq_enter_check_tick (); -/* Use the combo lockdep/tracing function */ -trace_hardirqs_off (); -instrumentation_end (); - -return ret; -} - -static void idtentry_exit_cond_resched (struct pt_regs *regs,bool may_sched) -{ -if (may_sched && !preempt_count ()) { -/* Sanity check RCU and thread stack */ -rcu_irq_exit_check_preempt (); -if (IS_ENABLED (CONFIG_DEBUG_ENTRY )) -WARN_ON_ONCE (!on_thread_stack ()); -if (need_resched ()) -preempt_schedule_irq (); - } -/* Covers both tracing and lockdep */ -trace_hardirqs_on (); -} - -/** -* idtentry_exit - Handle return from exception that used idtentry_enter() -* @regs: Pointer to pt_regs (exception entry regs) -* @state: Return value from matching call to idtentry_enter() -* -* Depending on the return target (kernel/user) this runs the necessary -* preemption and work checks if possible and reguired and returns to -* the caller with interrupts disabled and no further work pending. -* -* This is the last action before returning to the low level ASM code which -* just needs to return to the appropriate context. -* -* Counterpart to idtentry_enter(). The return value of the entry -* function must be fed into the @state argument. -*/ -void noinstr idtentry_exit(struct pt_regs *regs, idtentry_state_t state) -{ -lockdep_assert_irqs_disabled (); - -/* Check whether this returns to user mode */ -if (user_mode (regs)) { -irqentry_exit_to_user_mode (regs); - }else if (regs->flags &X86_EFLAGS_IF ) { -/* -* If RCU was not watching on entry this needs to be done -* carefully and needs the same ordering of lockdep/tracing -* and RCU as the return to user mode path. -*/ -if (state.exit_rcu) { -instrumentation_begin (); -/* Tell the tracer that IRET will enable interrupts */ -trace_hardirqs_on_prepare (); -lockdep_hardirqs_on_prepare (CALLER_ADDR0 ); -instrumentation_end (); -rcu_irq_exit (); -lockdep_hardirqs_on (CALLER_ADDR0 ); -return ; - } - -instrumentation_begin (); -idtentry_exit_cond_resched (regs,IS_ENABLED (CONFIG_PREEMPTION )); -instrumentation_end (); - }else { -/* -* IRQ flags state is correct already. Just tell RCU if it -* was not watching on entry. -*/ -if (state.exit_rcu) -rcu_irq_exit (); - } -} -#ifdef CONFIG_XEN_PV #ifndef CONFIG_PREEMPTION /*
@@ -427,7 +262,7 @@ __visible noinstr void xen_pv_evtchn_do_upcall(struct pt_regs *regs) inhcall =get_and_clear_inhcall ();if (inhcall && !WARN_ON_ONCE (state.exit_rcu)) {instrumentation_begin (); -idtentry_exit_cond_resched (regs, true); +irqentry_exit_cond_resched ();instrumentation_end ();restore_inhcall (inhcall); }else {
arch/x86/include/asm/idtentry.h +4 -6
@@ -11,12 +11,10 @@#include <asm/irq_stack.h> -typedef struct idtentry_state { -bool exit_rcu; -}idtentry_state_t ; - -idtentry_state_t idtentry_enter (struct pt_regs *regs); -void idtentry_exit (struct pt_regs *regs,idtentry_state_t state); +/* Temporary defines */ +typedef irqentry_state_t idtentry_state_t ; +#define idtentry_enter irqentry_enter +#define idtentry_exit irqentry_exit/** * DECLARE_IDTENTRY - Declare functions for simple IDT entry points
PATCH v5 14/15 x86/entry: Cleanup idtentry_enter/exit
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Remove the temporary defines and fixup all references.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/entry/common.c | 6 +++--- arch/x86/include/asm/idtentry.h | 33 ++++++++++++++------------------- arch/x86/kernel/kvm.c | 6 +++--- arch/x86/kernel/traps.c | 6 +++--- arch/x86/mm/fault.c | 6 +++--- 5 files changed, 26 insertions(+), 31 deletions(-)
arch/x86/entry/common.c +3 -3
@@ -248,9 +248,9 @@ static void __xen_pv_evtchn_do_upcall(vo {struct pt_regs *old_regs;bool inhcall; -idtentry_state_t state; +irqentry_state_t state; - state =idtentry_enter (regs); + state =irqentry_enter (regs); old_regs =set_irq_regs (regs);instrumentation_begin ();
@@ -266,7 +266,7 @@ static void __xen_pv_evtchn_do_upcall(voinstrumentation_end (); restore_inhcall(inhcall); }else { -idtentry_exit (regs, state); +irqentry_exit (regs, state); } }#endif /* CONFIG_XEN_PV */
arch/x86/include/asm/idtentry.h +14 -19
@@ -11,11 +11,6 @@#include <asm/irq_stack.h> -/* Temporary defines */ -typedef irqentry_state_t idtentry_state_t ; -#define idtentry_enter irqentry_enter -#define idtentry_exit irqentry_exit -/** * DECLARE_IDTENTRY - Declare functions for simple IDT entry points * No error code pushed by hardware
@@ -45,8 +40,8 @@ typedef irqentry_state_t idtentry_state_* The macro is written so it acts as function definition. Append the * body with a pair of curly brackets. * -* idtentry_enter() contains common code which has to be invoked before -* arbitrary code in the body. idtentry_exit() contains common code +* irqentry_enter() contains common code which has to be invoked before +* arbitrary code in the body. irqentry_exit() contains common code * which has to run before returning to the low level assembly code. */ #define DEFINE_IDTENTRY (func) \
@@ -54,12 +49,12 @@ static __always_inline void __##func(str \ __visible noinstr void func(struct pt_regs *regs) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \ __##func (regs); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs)
@@ -101,12 +96,12 @@ static __always_inline void __##func(str __visible noinstr void func(struct pt_regs *regs, \unsigned long error_code) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \ __##func (regs, error_code); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs, \
@@ -161,7 +156,7 @@ static __always_inline void __##func(str* body with a pair of curly brackets. * * Contrary to DEFINE_IDTENTRY_ERRORCODE() this does not invoke the -* idtentry_enter/exit() helpers before and after the body invocation. This +* irqentry_enter/exit() helpers before and after the body invocation. This * needs to be done in the body itself if applicable. Use if extra work * is required before the enter/exit() helpers are invoked. */
@@ -197,7 +192,7 @@ static __always_inline void __##func(str __visible noinstr void func(struct pt_regs *regs, \unsigned long error_code) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \irq_enter_rcu (); \
@@ -205,7 +200,7 @@ static __always_inline void __##func(str __##func (regs, (u8)error_code); \ irq_exit_rcu(); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs, u8 vector)
@@ -229,7 +224,7 @@ static __always_inline void __##func(str* DEFINE_IDTENTRY_SYSVEC - Emit code for system vector IDT entry points * @func: Function name of the entry point * -* idtentry_enter/exit() and irq_enter/exit_rcu() are invoked before the +* irqentry_enter/exit() and irq_enter/exit_rcu() are invoked before the * function body. KVM L1D flush request is set. * * Runs the function on the interrupt stack if the entry hit kernel mode
@@ -239,7 +234,7 @@ static void __##func(struct pt_regs *reg \ __visible noinstr void func(struct pt_regs *regs) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \irq_enter_rcu (); \
@@ -247,7 +242,7 @@ static void __##func(struct pt_regs *reg run_on_irqstack_cond(__##func, regs, regs); \ irq_exit_rcu(); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static noinline void __##func(struct pt_regs *regs)
@@ -268,7 +263,7 @@ static __always_inline void __##func(str \ __visible noinstr void func(struct pt_regs *regs) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \__irq_enter_raw (); \
@@ -276,7 +271,7 @@ static __always_inline void __##func(str __##func (regs); \ __irq_exit_raw(); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs)
arch/x86/kernel/kvm.c +3 -3
@@ -233,7 +233,7 @@ EXPORT_SYMBOL_GPL(kvm_read_and_reset_apf noinstr bool __kvm_handle_async_pf(struct pt_regs *regs, u32 token) {u32 reason =kvm_read_and_reset_apf_flags (); -idtentry_state_t state; +irqentry_state_t state;switch (reason) {case KVM_PV_REASON_PAGE_NOT_PRESENT :
@@ -243,7 +243,7 @@ noinstr bool __kvm_handle_async_pf(strucreturn false; } - state =idtentry_enter (regs); + state =irqentry_enter (regs);instrumentation_begin ();/*
@@ -264,7 +264,7 @@ noinstr bool __kvm_handle_async_pf(struc }instrumentation_end (); -idtentry_exit (regs, state); +irqentry_exit (regs, state);return true; }
arch/x86/kernel/traps.c +3 -3
@@ -245,7 +245,7 @@ static noinstr bool handle_bug(struct ptDEFINE_IDTENTRY_RAW (exc_invalid_op) { -idtentry_state_t state; +irqentry_state_t state;/* * We use UD2 as a short encoding for 'CALL __WARN', as such
@@ -255,11 +255,11 @@ DEFINE_IDTENTRY_RAW(exc_invalid_op)if (!user_mode (regs) &&handle_bug (regs))return ; - state =idtentry_enter (regs); + state =irqentry_enter (regs);instrumentation_begin ();handle_invalid_op (regs);instrumentation_end (); -idtentry_exit (regs, state); +irqentry_exit (regs, state); }DEFINE_IDTENTRY (exc_coproc_segment_overrun)
arch/x86/mm/fault.c +3 -3
@@ -1377,7 +1377,7 @@ handle_page_fault(struct pt_regs *regs,DEFINE_IDTENTRY_RAW_ERRORCODE (exc_page_fault) {unsigned long address =read_cr2 (); -idtentry_state_t state; +irqentry_state_t state;prefetchw (¤t->mm->mmap_lock);
@@ -1412,11 +1412,11 @@ DEFINE_IDTENTRY_RAW_ERRORCODE(exc_page_f* code reenabled RCU to avoid subsequent wreckage which helps * debugability. */ - state =idtentry_enter (regs); + state =irqentry_enter (regs);instrumentation_begin ();handle_page_fault (regs, error_code, address);instrumentation_end (); -idtentry_exit (regs, state); +irqentry_exit (regs, state); }
[tip: x86/entry] x86/entry: Cleanup idtentry_enter/exit
The following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: a27a0a55495cdde4b8d98f82460dc46eb44777fd Gitweb: https://git.kernel.org/tip/a27a0a55495cdde4b8d98f82460dc46eb44777fd Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:08 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:05:01 +02:00
x86/entry: Cleanup idtentry_enter/exit
Remove the temporary defines and fixup all references.
Signed-off-by Thomas Gleixner <[email protected]>
Reviewed-by Kees Cook <[email protected]>
--- arch/x86/entry/common.c | 6 +++--- arch/x86/include/asm/idtentry.h | 33 +++++++++++++------------------- arch/x86/kernel/kvm.c | 6 +++--- arch/x86/kernel/traps.c | 6 +++--- arch/x86/mm/fault.c | 6 +++--- 5 files changed, 26 insertions(+), 31 deletions(-)
arch/x86/entry/common.c +3 -3
@@ -248,9 +248,9 @@ __visible noinstr void xen_pv_evtchn_do_upcall(struct pt_regs *regs) {struct pt_regs *old_regs;bool inhcall; -idtentry_state_t state; +irqentry_state_t state; - state =idtentry_enter (regs); + state =irqentry_enter (regs); old_regs =set_irq_regs (regs);instrumentation_begin ();
@@ -266,7 +266,7 @@ __visible noinstr void xen_pv_evtchn_do_upcall(struct pt_regs *regs)instrumentation_end (); restore_inhcall(inhcall); }else { -idtentry_exit (regs, state); +irqentry_exit (regs, state); } }#endif /* CONFIG_XEN_PV */
arch/x86/include/asm/idtentry.h +14 -19
@@ -11,11 +11,6 @@#include <asm/irq_stack.h> -/* Temporary defines */ -typedef irqentry_state_t idtentry_state_t ; -#define idtentry_enter irqentry_enter -#define idtentry_exit irqentry_exit -/** * DECLARE_IDTENTRY - Declare functions for simple IDT entry points * No error code pushed by hardware
@@ -45,8 +40,8 @@ typedef irqentry_state_t idtentry_state_t;* The macro is written so it acts as function definition. Append the * body with a pair of curly brackets. * -* idtentry_enter() contains common code which has to be invoked before -* arbitrary code in the body. idtentry_exit() contains common code +* irqentry_enter() contains common code which has to be invoked before +* arbitrary code in the body. irqentry_exit() contains common code * which has to run before returning to the low level assembly code. */ #define DEFINE_IDTENTRY (func) \
@@ -54,12 +49,12 @@ static __always_inline void __##func(struct pt_regs *regs); \ \ __visible noinstr void func(struct pt_regs *regs) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \ __##func (regs); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs)
@@ -101,12 +96,12 @@ static __always_inline void __##func(struct pt_regs *regs, \ __visible noinstr void func(struct pt_regs *regs, \unsigned long error_code) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \ __##func (regs, error_code); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs, \
@@ -161,7 +156,7 @@ __visible noinstr void func(struct pt_regs *regs)* body with a pair of curly brackets. * * Contrary to DEFINE_IDTENTRY_ERRORCODE() this does not invoke the -* idtentry_enter/exit() helpers before and after the body invocation. This +* irqentry_enter/exit() helpers before and after the body invocation. This * needs to be done in the body itself if applicable. Use if extra work * is required before the enter/exit() helpers are invoked. */
@@ -197,7 +192,7 @@ static __always_inline void __##func(struct pt_regs *regs, u8 vector); \ __visible noinstr void func(struct pt_regs *regs, \unsigned long error_code) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \irq_enter_rcu (); \
@@ -205,7 +200,7 @@ __visible noinstr void func(struct pt_regs *regs, \ __##func (regs, (u8)error_code); \ irq_exit_rcu(); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs, u8 vector)
@@ -229,7 +224,7 @@ static __always_inline void __##func(struct pt_regs *regs, u8 vector)* DEFINE_IDTENTRY_SYSVEC - Emit code for system vector IDT entry points * @func: Function name of the entry point * -* idtentry_enter/exit() and irq_enter/exit_rcu() are invoked before the +* irqentry_enter/exit() and irq_enter/exit_rcu() are invoked before the * function body. KVM L1D flush request is set. * * Runs the function on the interrupt stack if the entry hit kernel mode
@@ -239,7 +234,7 @@ static void __##func(struct pt_regs *regs); \ \ __visible noinstr void func(struct pt_regs *regs) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \irq_enter_rcu (); \
@@ -247,7 +242,7 @@ __visible noinstr void func(struct pt_regs *regs) \ run_on_irqstack_cond(__##func, regs, regs); \ irq_exit_rcu(); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static noinline void __##func(struct pt_regs *regs)
@@ -268,7 +263,7 @@ static __always_inline void __##func(struct pt_regs *regs); \ \ __visible noinstr void func(struct pt_regs *regs) \ { \ -idtentry_state_t state =idtentry_enter (regs); \ +irqentry_state_t state =irqentry_enter (regs); \ \instrumentation_begin (); \__irq_enter_raw (); \
@@ -276,7 +271,7 @@ __visible noinstr void func(struct pt_regs *regs) \ __##func (regs); \ __irq_exit_raw(); \ instrumentation_end(); \ - idtentry_exit(regs, state); \ + irqentry_exit(regs, state); \ } \ \ static __always_inline void __##func(struct pt_regs *regs)
arch/x86/kernel/kvm.c +3 -3
@@ -233,7 +233,7 @@ EXPORT_SYMBOL_GPL(kvm_read_and_reset_apf_flags); noinstr bool __kvm_handle_async_pf(struct pt_regs *regs, u32 token) {u32 reason =kvm_read_and_reset_apf_flags (); -idtentry_state_t state; +irqentry_state_t state;switch (reason) {case KVM_PV_REASON_PAGE_NOT_PRESENT :
@@ -243,7 +243,7 @@ noinstr bool __kvm_handle_async_pf(struct pt_regs *regs, u32 token)return false; } - state =idtentry_enter (regs); + state =irqentry_enter (regs);instrumentation_begin ();/*
@@ -264,7 +264,7 @@ noinstr bool __kvm_handle_async_pf(struct pt_regs *regs, u32 token) }instrumentation_end (); -idtentry_exit (regs, state); +irqentry_exit (regs, state);return true; }
arch/x86/kernel/traps.c +3 -3
@@ -245,7 +245,7 @@ static noinstr bool handle_bug(struct pt_regs *regs)DEFINE_IDTENTRY_RAW (exc_invalid_op) { -idtentry_state_t state; +irqentry_state_t state;/* * We use UD2 as a short encoding for 'CALL __WARN', as such
@@ -255,11 +255,11 @@ DEFINE_IDTENTRY_RAW(exc_invalid_op)if (!user_mode (regs) &&handle_bug (regs))return ; - state =idtentry_enter (regs); + state =irqentry_enter (regs);instrumentation_begin ();handle_invalid_op (regs);instrumentation_end (); -idtentry_exit (regs, state); +irqentry_exit (regs, state); }DEFINE_IDTENTRY (exc_coproc_segment_overrun)
arch/x86/mm/fault.c +3 -3
@@ -1377,7 +1377,7 @@ handle_page_fault(struct pt_regs *regs, unsigned long error_code,DEFINE_IDTENTRY_RAW_ERRORCODE (exc_page_fault) {unsigned long address =read_cr2 (); -idtentry_state_t state; +irqentry_state_t state;prefetchw (¤t->mm->mmap_lock);
@@ -1412,11 +1412,11 @@ DEFINE_IDTENTRY_RAW_ERRORCODE(exc_page_fault)* code reenabled RCU to avoid subsequent wreckage which helps * debugability. */ - state =idtentry_enter (regs); + state =irqentry_enter (regs);instrumentation_begin ();handle_page_fault (regs, error_code, address);instrumentation_end (); -idtentry_exit (regs, state); +irqentry_exit (regs, state); }
PATCH v5 15/15 x86/kvm: Use generic xfer to guest work function
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Use the generic infrastructure to check for and handle pending work before transitioning into guest mode.
This now handles TIF_NOTIFY_RESUME as well which was ignored so far. Handling it is important as this covers task work and task work will be used to offload the heavy lifting of POSIX CPU timers to thread context.
Signed-off-by Thomas Gleixner <[email protected]>
---
V5: Rename exit -> xfer
--- arch/x86/kvm/Kconfig | 1 + arch/x86/kvm/vmx/vmx.c | 11 +++++------ arch/x86/kvm/x86.c | 15 ++++++--------- 3 files changed, 12 insertions(+), 15 deletions(-)
arch/x86/kvm/Kconfig +1 -0
@@ -42,6 +42,7 @@ config KVM select HAVE_KVM_MSI select HAVE_KVM_CPU_RELAX_INTERCEPT select HAVE_KVM_NO_POLL + select KVM_XFER_TO_GUEST_WORK select KVM_GENERIC_DIRTYLOG_READ_PROTECT select KVM_VFIO select SRCU
arch/x86/kvm/vmx/vmx.c +5 -6
@@ -27,6 +27,7 @@#include <linux/slab.h> #include <linux/tboot.h> #include <linux/trace_events.h> +#include <linux/entry-kvm.h> #include <asm/apic.h> #include <asm/asm.h>
@@ -5376,14 +5377,12 @@ static int handle_invalid_guest_state(st }/* -* Note, return 1 and not 0, vcpu_run() is responsible for -* morphing the pending signal into the proper return code. +* Note, return 1 and not 0, vcpu_run() will invoke +* xfer_to_guest_mode() which will create a proper return +* code. */ -if (signal_pending (current)) +if (__xfer_to_guest_mode_work_pending ())return 1 ; - -if (need_resched ()) -schedule (); }return 1 ;
arch/x86/kvm/x86.c +6 -9
@@ -56,6 +56,7 @@#include <linux/sched/stat.h> #include <linux/sched/isolation.h> #include <linux/mem_encrypt.h> +#include <linux/entry-kvm.h> #include <trace/events/kvm.h>
@@ -1585,7 +1586,7 @@ EXPORT_SYMBOL_GPL(kvm_emulate_wrmsr);bool kvm_vcpu_exit_request (struct kvm_vcpu *vcpu) {return vcpu->mode ==EXITING_GUEST_MODE ||kvm_request_pending (vcpu) || -need_resched () ||signal_pending (current); +xfer_to_guest_mode_work_pending (); }EXPORT_SYMBOL_GPL (kvm_vcpu_exit_request);
@@ -8676,15 +8677,11 @@ static int vcpu_run(struct kvm_vcpu *vcpbreak ; } -if (signal_pending (current)) { - r = -EINTR ; - vcpu->run->exit_reason =KVM_EXIT_INTR ; - ++vcpu->stat.signal_exits; -break ; - } -if (need_resched ()) { +if (xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); -cond_resched (); + r =xfer_to_guest_mode (vcpu); +if (r) +return r; vcpu->srcu_idx =srcu_read_lock (&kvm->srcu); } }
Re: [patch V5 15/15] x86/kvm: Use generic xfer to guest work function
Quoted 5 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Use the generic infrastructure to check for and handle pending work before transitioning into guest mode.
Quoted 8 lines
This now handles
TIF_NOTIFY_RESUMEas well which was ignored so far. Handling it is important as this covers task work and task work will be used to offload the heavy lifting of POSIX CPU timers to thread context.Signed-off-by Thomas Gleixner <[email protected]>---
V5: Rename exit -> xfer
---
One nit/question below (though it's really about patch 5).
Reviewed-and-tested-by Sean Christopherson <[email protected]>
Quoted 5 lines
arch/x86/kvm/x86.c +0
-1@@ -8676,15 +8677,11 @@ static int vcpu_run(struct kvm_vcpu *vcpbreak ; } -if (signal_pending (current)) {
arch/x86/kvm/x86.c static int vcpu_run(struct kvm_vcpu *vcp · line 8682
Quoted 10 lines
arch/x86/kvm/x86.c +2
-7- r = -EINTR ; - vcpu->run->exit_reason =KVM_EXIT_INTR ; - ++vcpu->stat.signal_exits; -break ; - } -if (need_resched ()) { +if (xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); -cond_resched (); + r =xfer_to_guest_mode (vcpu);
Any reason not to call this xfer_to_guest_mode_work()? Or handle_work(), do_work(), etc... Without the "work" part, it looks like a function that should be invoked unconditionally. It's obvious that's not the case if one looks at the implementation, but it's a bit confusing on the KVM side of things.
Quoted 6 lines
arch/x86/kvm/x86.c +2
-0+if (r) +return r; vcpu->srcu_idx =srcu_read_lock (&kvm->srcu); } }
Re: [patch V5 15/15] x86/kvm: Use generic xfer to guest work function
Sean,
Quoted 11 lines · Sean Christopherson
arch/x86/kvm/x86.c static int vcpu_run(struct kvm_vcpu *vcp · line 8682
Quoted 4 lines · Thomas Gleixner
arch/x86/kvm/x86.c +2
-1+if (xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); -cond_resched (); + r =xfer_to_guest_mode (vcpu);Any reason not to call this
xfer_to_guest_mode_work()? Orhandle_work(),do_work(), etc... Without the "work" part, it looks like a function that should be invoked unconditionally. It's obvious that's not the case if one looks at the implementation, but it's a bit confusing on the KVM side of things.
The reason is probably lazyness. The original approach was to have this as close as possible to user entry/exit but with the recent changes vs. instrumentation and RCU this is not longer the case.
I really want to keep the notion of transitioning in the function name, so xfer_to_guest_mode_handle_work() makes a lot of sense.
I'll change that before merging the lot into the tip tree if your Reviewed-by still stands with that change made w/o reposting.
Thanks,
tglx
Re: [patch V5 15/15] x86/kvm: Use generic xfer to guest work function
Quoted 19 lines · Thomas Gleixner
Sean,
Quoted 11 lines · Sean Christopherson
arch/x86/kvm/x86.c static int vcpu_run(struct kvm_vcpu *vcp · line 8682
Quoted 4 lines · Thomas Gleixner
arch/x86/kvm/x86.c +2
-1+if (xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); -cond_resched (); + r =xfer_to_guest_mode (vcpu);Any reason not to call this
xfer_to_guest_mode_work()? Orhandle_work(),do_work(), etc... Without the "work" part, it looks like a function that should be invoked unconditionally. It's obvious that's not the case if one looks at the implementation, but it's a bit confusing on the KVM side of things.The reason is probably lazyness. The original approach was to have this as close as possible to user entry/exit but with the recent changes vs. instrumentation and RCU this is not longer the case.
Quoted 5 lines
I really want to keep the notion of transitioning in the function name, so
xfer_to_guest_mode_handle_work()makes a lot of sense.I'll change that before merging the lot into the tip tree if your Reviewed-by still stands with that change made w/o reposting.
Ya, works for me.
Re: [patch V5 15/15] x86/kvm: Use generic xfer to guest work function
In reply to Thomas GleixnerQuoted 71 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Use the generic infrastructure to check for and handle pending work before transitioning into guest mode.
This now handles
TIF_NOTIFY_RESUMEas well which was ignored so far. Handling it is important as this covers task work and task work will be used to offload the heavy lifting of POSIX CPU timers to thread context.Signed-off-by Thomas Gleixner <[email protected]>---
V5: Rename exit -> xfer--- arch/x86/kvm/Kconfig | 1 + arch/x86/kvm/vmx/vmx.c | 11 +++++------ arch/x86/kvm/x86.c | 15 ++++++--------- 3 files changed, 12 insertions(+), 15 deletions(-)arch/x86/kvm/Kconfig +1
-0@@ -42,6 +42,7 @@ config KVM select HAVE_KVM_MSI select HAVE_KVM_CPU_RELAX_INTERCEPT select HAVE_KVM_NO_POLL + select KVM_XFER_TO_GUEST_WORK select KVM_GENERIC_DIRTYLOG_READ_PROTECT select KVM_VFIO select SRCUarch/x86/kvm/vmx/vmx.c +1
-0@@ -27,6 +27,7 @@#include <linux/slab.h> #include <linux/tboot.h> #include <linux/trace_events.h> +#include <linux/entry-kvm.h> #include <asm/apic.h> #include <asm/asm.h> @@ -5376,14 +5377,12 @@ static int handle_invalid_guest_state(streturn 1 ;arch/x86/kvm/x86.c +2
-7@@ -56,6 +56,7 @@#include <linux/sched/stat.h> #include <linux/sched/isolation.h> #include <linux/mem_encrypt.h> +#include <linux/entry-kvm.h> #include <trace/events/kvm.h> @@ -1585,7 +1586,7 @@ EXPORT_SYMBOL_GPL(kvm_emulate_wrmsr);bool kvm_vcpu_exit_request (struct kvm_vcpu *vcpu) {return vcpu->mode ==EXITING_GUEST_MODE ||kvm_request_pending (vcpu) || -need_resched () ||signal_pending (current); +xfer_to_guest_mode_work_pending (); }EXPORT_SYMBOL_GPL (kvm_vcpu_exit_request);@@ -8676,15 +8677,11 @@ static int vcpu_run(struct kvm_vcpu *vcpbreak ; } -if (signal_pending (current)) { - r = -EINTR ; - vcpu->run->exit_reason =KVM_EXIT_INTR ; - ++vcpu->stat.signal_exits; -break ; - }
arch/x86/kvm/x86.c static int vcpu_run(struct kvm_vcpu *vcp · line 8687
Quoted 10 lines
arch/x86/kvm/x86.c +4
-2-if (need_resched ()) { +if (xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); -cond_resched (); + r =xfer_to_guest_mode (vcpu); +if (r) +return r; vcpu->srcu_idx =srcu_read_lock (&kvm->srcu); } }
So this chunk replaces vcpu_run()'s cond_resched() call with a call to xfer_to_guest_mode(), which checks NEED_RESCHED & acts upon it, among other things.
But:
Quoted 6 lines
arch/x86/kvm/x86.c +1
-2}/* -* Note, return 1 and not 0, vcpu_run() is responsible for -* morphing the pending signal into the proper return code. +* Note, return 1 and not 0, vcpu_run() will invoke
arch/x86/kvm/x86.c
Quoted 10 lines
arch/x86/kvm/x86.c +3
-4+* xfer_to_guest_mode() which will create a proper return +* code. */ -if (signal_pending (current)) +if (__xfer_to_guest_mode_work_pending ())return 1 ; - -if (need_resched ()) -schedule (); }
AFAICS this chunk removes a conditional reschedule point from handle_invalid_guest_state() and replaces it with __xfer_to_guest_mode_work_pending().
But __xfer_to_guest_mode_work_pending() doesn't do the cond-resched of the full xfer_to_guest_mode_work() function - so we essentially lose a cond_resched() here.
Is this side effect intended, was the cond_resched() superfluous?
The logic is quite a maze, and the KVM code was missing a few of the state checks, so maybe this is all for the better - just wanted to mention it, in case it matters.
Thanks,
Ingo
Re: [patch V5 15/15] x86/kvm: Use generic xfer to guest work function
Quoted 20 lines · Ingo Molnar
Quoted 4 lines · Thomas Gleixner
arch/x86/kvm/vmx/vmx.c +1
-2/* -* Note, return 1 and not 0, vcpu_run() is responsible for -* morphing the pending signal into the proper return code. +* Note, return 1 and not 0, vcpu_run() will invoke arch/x86/kvm/vmx/vmx.c static int handle_invalid_guest_state(st · line 5386
Quoted 10 lines
arch/x86/kvm/vmx/vmx.c +3
-4+* xfer_to_guest_mode() which will create a proper return +* code. */ -if (signal_pending (current)) +if (__xfer_to_guest_mode_work_pending ())return 1 ; - -if (need_resched ()) -schedule (); }AFAICS this chunk removes a conditional reschedule point from
handle_invalid_guest_state()and replaces it with__xfer_to_guest_mode_work_pending().
Quoted 5 lines
But
__xfer_to_guest_mode_work_pending()doesn't do the cond-resched of the fullxfer_to_guest_mode_work()function - so we essentially lose acond_resched()here.Is this side effect intended, was the
cond_resched()superfluous?
It makes the thing drop back to the outer loop for any pending work not only for signals. That avoids having yet another thing to worry about.
Thanks,
tglx
[tip: x86/entry] x86/kvm: Use generic xfer to guest work function
In reply to Thomas GleixnerThe following commit has been merged into the x86/entry branch of tip:
The following commit has been merged into the x86/entry branch of tip:
Commit-ID: 72c3c0fe54a3f3ddea8f5ca468ddf9deaf2100b7 Gitweb: https://git.kernel.org/tip/72c3c0fe54a3f3ddea8f5ca468ddf9deaf2100b7 Author: Thomas Gleixner <[email protected]> AuthorDate: Thu, 23 Jul 2020 00:00:09 +02:00 Committer: Thomas Gleixner <[email protected]> CommitterDate: Fri, 24 Jul 2020 15:05:01 +02:00
x86/kvm: Use generic xfer to guest work function
Use the generic infrastructure to check for and handle pending work before transitioning into guest mode.
This now handles TIF_NOTIFY_RESUME as well which was ignored so far. Handling it is important as this covers task work and task work will be used to offload the heavy lifting of POSIX CPU timers to thread context.
Signed-off-by Thomas Gleixner <[email protected]>
--- arch/x86/kvm/Kconfig | 1 + arch/x86/kvm/vmx/vmx.c | 11 +++++------ arch/x86/kvm/x86.c | 15 ++++++--------- 3 files changed, 12 insertions(+), 15 deletions(-)
arch/x86/kvm/Kconfig +1 -0
@@ -42,6 +42,7 @@ config KVM select HAVE_KVM_MSI select HAVE_KVM_CPU_RELAX_INTERCEPT select HAVE_KVM_NO_POLL + select KVM_XFER_TO_GUEST_WORK select KVM_GENERIC_DIRTYLOG_READ_PROTECT select KVM_VFIO select SRCU
arch/x86/kvm/vmx/vmx.c +5 -6
@@ -27,6 +27,7 @@#include <linux/slab.h> #include <linux/tboot.h> #include <linux/trace_events.h> +#include <linux/entry-kvm.h> #include <asm/apic.h> #include <asm/asm.h>
@@ -5373,14 +5374,12 @@ static int handle_invalid_guest_state(struct kvm_vcpu *vcpu) }/* -* Note, return 1 and not 0, vcpu_run() is responsible for -* morphing the pending signal into the proper return code. +* Note, return 1 and not 0, vcpu_run() will invoke +* xfer_to_guest_mode() which will create a proper return +* code. */ -if (signal_pending (current)) +if (__xfer_to_guest_mode_work_pending ())return 1 ; - -if (need_resched ()) -schedule (); }return 1 ;
arch/x86/kvm/x86.c +6 -9
@@ -56,6 +56,7 @@#include <linux/sched/stat.h> #include <linux/sched/isolation.h> #include <linux/mem_encrypt.h> +#include <linux/entry-kvm.h> #include <trace/events/kvm.h>
@@ -1587,7 +1588,7 @@ EXPORT_SYMBOL_GPL(kvm_emulate_wrmsr);bool kvm_vcpu_exit_request (struct kvm_vcpu *vcpu) {return vcpu->mode ==EXITING_GUEST_MODE ||kvm_request_pending (vcpu) || -need_resched () ||signal_pending (current); +xfer_to_guest_mode_work_pending (); }EXPORT_SYMBOL_GPL (kvm_vcpu_exit_request);
@@ -8681,15 +8682,11 @@ static int vcpu_run(struct kvm_vcpu *vcpu)break ; } -if (signal_pending (current)) { - r = -EINTR ; - vcpu->run->exit_reason =KVM_EXIT_INTR ; - ++vcpu->stat.signal_exits; -break ; - } -if (need_resched ()) { +if (xfer_to_guest_mode_work_pending ()) {srcu_read_unlock (&kvm->srcu, vcpu->srcu_idx); -cond_resched (); + r =xfer_to_guest_mode_handle_work (vcpu); +if (r) +return r; vcpu->srcu_idx =srcu_read_lock (&kvm->srcu); } }
Quoted 2 lines · Thomas Gleixner
This is the 5th version of generic entry/exit functionality for host and guest.
I've merged the pile in two steps. Patch 1-5, i.e. the generic code is here:
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git core/entry
and merged this branch and patch 6-15 into
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86/entry
core/entry is immutable and any updates, changes go on top. It's meant as a base for other architecture developers who want to fiddle with that without having to get the x86 mess as well.
Thanks for all the help!
tglx
Quoted 9 lines · Thomas Gleixner
Quoted 2 lines · Thomas Gleixner
This is the 5th version of generic entry/exit functionality for host and guest.
I've merged the pile in two steps. Patch 1-5, i.e. the generic code is here:
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.gitcore/entry
Quoted 7 lines
and merged this branch and patch 6-15 into
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.gitx86/entrycore/entry is immutable and any updates, changes go on top. It's meant as a base for other architecture developers who want to fiddle with that without having to get the x86 mess as well.
Hi Thomas,
Just FYI: I'm going to take a look at arm64 support for this. I know Mark previously posted[1] a series moving much of the code to C, so hopefully I can build on that and make use of the generic code.
Thanks,
Steve