- 11 Jun, 2020 40 commits
-
-
Thomas Gleixner authored
Convert the XEN/PV hypercall to IDTENTRY: - Emit the ASM stub with DECLARE_IDTENTRY - Remove the ASM idtentry in 64-bit - Remove the open coded ASM entry code in 32-bit - Remove the old prototypes The handler stubs need to stay in ASM code as they need corner case handling and adjustment of the stack pointer. Provide a new C function which invokes the entry/exit handling and calls into the XEN handler on the interrupt stack if required. The exit code is slightly different from the regular idtentry_exit() on non-preemptible kernels. If the hypercall is preemptible and need_resched() is set then XEN provides a preempt hypercall scheduling function. Move this functionality into the entry code so it can use the existing idtentry functionality. [ mingo: Build fixes. ] Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
Ingo Molnar <mingo@kernel.org> Acked-by:
Andy Lutomirski <luto@kernel.org> Acked-by:
Juergen Gross <jgross@suse.com> Tested-by:
-
Thomas Gleixner authored
The XEN PV hypercall requires the ability of conditional rescheduling when preemption is disabled because some hypercalls take ages. Split out the rescheduling code from idtentry_exit_cond_rcu() so it can be reused for that. Signed-off-by:
-
Thomas Gleixner authored
The first step to get rid of the ENTER/LEAVE_IRQ_STACK ASM macro maze. Use the new C code helpers to move do_softirq_own_stack() out of ASM code. Signed-off-by:
-
Thomas Gleixner authored
Device interrupt handlers and system vector handlers are executed on the interrupt stack. The stack switch happens in the low level assembly entry code. This conflicts with the efforts to consolidate the exit code in C to ensure correctness vs. RCU and tracing. As there is no way to move #DB away from IST due to the MOV SS issue, the requirements vs. #DB and NMI for switching to the interrupt stack do not exist anymore. The only requirement is that interrupts are disabled. That allows the moving of the stack switching to C code, which simplifies the entry/exit handling further, because it allows the switching of stacks after handling the entry and on exit before handling RCU, returning to usermode and kernel preemption in the same way as for regular exceptions. The initial attempt of having the stack switching in inline ASM caused too much headache vs. objtool and the unwinder. After analysing the use cases it was agreed on that having the stack switch in ASM for the price of an indirect call is acceptable, as the main users are indirect call heavy anyway and the few system vectors which are empty shells (scheduler IPI and KVM posted interrupt vectors) can run from the regular stack. Provide helper functions to check whether the interrupt stack is already active and whether stack switching is required. 64-bit only for now, as 32-bit has a variant of that already. Once this is cleaned up, the two implementations might be consolidated as an additional cleanup on top. Signed-off-by:
-
Thomas Gleixner authored
Like __irq_enter/exit() but without time accounting. To be used for "empty" system vectors like the scheduler IPI to avoid the overhead. Signed-off-by:
-
Thomas Gleixner authored
irq_enter()/exit() currently include RCU handling. To properly separate the RCU handling code, provide variants which contain only the non-RCU related functionality. Signed-off-by:
-
Thomas Gleixner authored
Now that everything is converted to conditional RCU handling remove idtentry_enter/exit() and tidy up the conditional functions. This does not remove rcu_irq_exit_preempt(), to avoid conflicts with the RCU tree. Will be removed once all of this hits Linus's tree. Signed-off-by:
-
Thomas Gleixner authored
Switch all idtentry_enter/exit() users over to the new conditional RCU handling scheme and make the user mode entries in #DB, #INT3 and #MCE use the user mode idtentry functions. Signed-off-by:
-
Thomas Gleixner authored
As there are exceptions which already handle entry from user mode and from kernel mode separately, providing explicit user entry/exit handling callbacks makes sense and makes the code easier to understand. Signed-off-by:
-
Thomas Gleixner authored
After a lengthy discussion [1] it turned out that RCU does not need a full rcu_irq_enter/exit() when RCU is already watching. All it needs if NOHZ_FULL is active is to check whether the tick needs to be restarted. This allows to avoid a separate variant for the pagefault handler which cannot invoke rcu_irq_enter() on a kernel pagefault which might sleep. The cond_rcu argument is only temporary and will be removed once the existing users of idtentry_enter/exit() have been cleaned up. After that the code can be significantly simplified. [ mingo: Simplified the control flow ] Signed-off-by:
"Paul E. McKenney" <paulmck@kernel.org> Acked-by:
-
Thomas Gleixner authored
The hardware latency tracer calls into instrumentable functions. Move the calls into the RCU watching sections and annotate them. Signed-off-by:
-
Ingo Molnar authored
The following commit: 095b7a3e7745 ("x86/entry: Convert double fault exception to IDTENTRY_DF") introduced a new build warning on 64-bit allnoconfig kernels, that have CONFIG_VMAP_STACK disabled: arch/x86/kernel/traps.c:332:16: warning: unused variable ‘address’ [-Wunused-variable] This variable is only used if CONFIG_VMAP_STACK is defined, so make it dependent on that, not CONFIG_X86_64. Signed-off-by: -
Thomas Gleixner authored
Convert #DF to IDTENTRY_DF - Implement the C entry point with DEFINE_IDTENTRY_DF - Emit the ASM stub with DECLARE_IDTENTRY_DF on 64bit - Remove the ASM idtentry in 64bit - Adjust the 32bit shim code - Fixup the XEN/PV code - Remove the old prototypes No functional change. Signed-off-by:
Alexandre Chartre <alexandre.chartre@oracle.com> Acked-by:
Peter Zijlstra <peterz@infradead.org> Acked-by:
-
Thomas Gleixner authored
Provide a separate macro for #DF as this needs to emit paranoid only code and has also a special ASM stub in 32bit. Signed-off-by:
-
Thomas Gleixner authored
Mark the relevant functions noinstr, use the plain non-instrumented MSR accessors. The only odd part is the instrumentation_begin()/end() pair around the indirect machine_check_vector() call as objtool can't figure that out. The possible invoked functions are annotated correctly. Also use notrace variant of nmi_enter/exit(). If MCEs happen then hardware latency tracing is the least of the worries. Signed-off-by:
-
Thomas Gleixner authored
The functions invoked from handle_debug() can be instrumented. Tell objtool about it. Signed-off-by:
-
Thomas Gleixner authored
Now that there are separate entry points, move the kernel/user_mode specifc checks into the entry functions so the common handling code does not need the extra mode checks. Make the code more readable while at it. Signed-off-by:
-
Thomas Gleixner authored
The MCE entry point uses the same mechanism as the IST entry point for now. For #DB split the inner workings and just keep the nmi_enter/exit() magic in the IST variant. Fixup the ASM code to emit the proper noist_##cfunc call. Signed-off-by:
-
Thomas Gleixner authored
Provide NOIST entry point macros which allows to implement NOIST variants of the C entry points. These are invoked when #DB or #MC enter from user space. This allows explicit handling of the difference between user mode and kernel mode entry later. Signed-off-by:
-
Thomas Gleixner authored
The C entry points do not expect an error code. Signed-off-by:
-
Thomas Gleixner authored
Convert #DB to IDTENTRY_ERRORCODE: - Implement the C entry point with DEFINE_IDTENTRY_DB - Emit the ASM stub with DECLARE_IDTENTRY - Remove the ASM idtentry in 64bit - Remove the open coded ASM entry code in 32bit - Fixup the XEN/PV code - Remove the old prototypes No functional change. Signed-off-by:
-
Peter Zijlstra authored
DR6/7 should be handled before nmi_enter() is invoked and restore after nmi_exit() to minimize the exposure. Split it out into helper inlines and bring it into the correct order. Signed-off-by:
-
Thomas Gleixner authored
Mark all functions in the fragile code parts noinstr or force inlining so they can't be instrumented. Also make the hardware latency tracer invocation explicit outside of non-instrumentable section. Signed-off-by:
-
Thomas Gleixner authored
Convert #NMI to IDTENTRY_NMI: - Implement the C entry point with DEFINE_IDTENTRY_NMI - Fixup the XEN/PV code - Remove the old prototypes No functional change. Signed-off-by:
-
Thomas Gleixner authored
XEN/PV has special wrappers for NMI and DB exceptions. They redirect these exceptions through regular IDTENTRY points. Provide the necessary IDTENTRY macros to make this work Signed-off-by:
-
Thomas Gleixner authored
mce_check_crashing_cpu() is called right at the entry of the MCE handler. It uses mce_rdmsr() and mce_wrmsr() which are wrappers around rdmsr() and wrmsr() to handle the MCE error injection mechanism, which is pointless in this context, i.e. when the MCE hits an offline CPU or the system is already marked crashing. The MSR access can also be traced, so use the untraceable variants. This is also safe vs. XEN paravirt as these MSRs are not affected by XEN PV modifications. Signed-off-by:
-
Thomas Gleixner authored
Convert #MC to IDTENTRY_MCE: - Implement the C entry points with DEFINE_IDTENTRY_MCE - Emit the ASM stub with DECLARE_IDTENTRY_MCE - Remove the ASM idtentry in 64bit - Remove the open coded ASM entry code in 32bit - Fixup the XEN/PV code - Remove the old prototypes - Remove the error code from *machine_check_vector() as it is always 0 and not used by any of the functions it can point to. Fixup all the functions as well. No functional change. Signed-off-by: -
Thomas Gleixner authored
There is no reason to have nmi_enter/exit() in the actual MCE handlers. Move it to the entry point. This also covers the until now uncovered initial handler which only prints. Signed-off-by:
-
Thomas Gleixner authored
Same as IDTENTRY but for exceptions which run on Interrupt Stacks (IST) on 64bit. For 32bit this maps to IDTENTRY. There are 3 variants which will be used: IDTENTRY_MCE IDTENTRY_DB IDTENTRY_NMI These map to IDTENTRY_IST, but only the MCE and DB variants are emitting ASM code as the NMI entry needs hand crafted ASM still. The function defines do not contain any idtenter/exit calls as these exceptions need special treatment. Signed-off-by: -
Peter Zijlstra authored
For code simplicity split up the int3 handler into a kernel and user part which makes the code flow simpler to understand. Signed-off-by:
-
Thomas Gleixner authored
Convert #BP to IDTENTRY_RAW: - Implement the C entry point with DEFINE_IDTENTRY_RAW - Invoke idtentry_enter/exit() from the function body - Emit the ASM stub with DECLARE_IDTENTRY_RAW - Remove the ASM idtentry in 64bit - Remove the open coded ASM entry code in 32bit - Fixup the XEN/PV code - Remove the old prototypes No functional change. This could be a plain IDTENTRY, but as Peter pointed out INT3 is broken vs. the static key in the context tracking code as this static key might be in the state of being patched and has an int3 which would recurse forever. IDTENTRY_RAW is therefore chosen to allow addressing this issue without lots of code churn. Signed-off-by:
-
Thomas Gleixner authored
Some exception handlers need to do extra work before any of the entry helpers are invoked. Provide IDTENTRY_RAW for this. Signed-off-by:
-
Peter Zijlstra authored
Avoid calling out to bsearch() by inlining it, for normal kernel configs this was the last external call and poke_int3_handler() is now fully self sufficient -- no calls to external code. Signed-off-by:
-
Peter Zijlstra authored
For code that needs the ultimate performance (it can inline the @cmp function too) or simply needs to avoid calling external functions for whatever reason, provide an __always_inline variant of bsearch(). [ tglx: Renamed to __inline_bsearch() as suggested by Andy ] Signed-off-by:
-
Peter Zijlstra authored
Use arch_atomic_*() and __READ_ONCE() to ensure nothing untoward creeps in and ruins things. That is; this is the INT3 text poke handler, strictly limit the code that runs in it, lest it inadvertenly hits yet another INT3. Reported-by:
Masami Hiramatsu <mhiramat@kernel.org> Reviewed-by:
-
Thomas Gleixner authored
In order to ensure poke_int3_handler() is completely self contained -- this is called while modifying other text, imagine the fun of hitting another INT3 -- ensure that everything it uses is not traced. The primary means here is to force inlining; bsearch() is notrace because all of lib/ is. Signed-off-by:
-
Thomas Gleixner authored
Convert the IRET exception handler to IDTENTRY_SW. This is slightly different than the conversions of hardware exceptions as the IRET exception is invoked via an exception table when IRET faults. So it just uses the IDTENTRY_SW mechanism for consistency. It does not emit ASM code as it does not fit the other idtentry exceptions. - Implement the C entry point with DEFINE_IDTENTRY_SW() which maps to DEFINE_IDTENTRY() - Fixup the XEN/PV code - Remove the old prototypes - Remove the RCU warning as the new entry macro ensures correctness No functional change. Signed-off-by: -
Thomas Gleixner authored
Convert #XF to IDTENTRY_ERRORCODE: - Implement the C entry point with DEFINE_IDTENTRY - Emit the ASM stub with DECLARE_IDTENTRY - Handle INVD_BUG in C - Remove the ASM idtentry in 64bit - Remove the open coded ASM entry code in 32bit - Fixup the XEN/PV code - Remove the old prototypes - Remove the RCU warning as the new entry macro ensures correctness No functional change. Signed-off-by:
-
Thomas Gleixner authored
Convert #AC to IDTENTRY_ERRORCODE: - Implement the C entry point with DEFINE_IDTENTRY - Emit the ASM stub with DECLARE_IDTENTRY - Remove the ASM idtentry in 64bit - Remove the open coded ASM entry code in 32bit - Fixup the XEN/PV code - Remove the old prototypes - Remove the RCU warning as the new entry macro ensures correctness No functional change. Signed-off-by:
-
Thomas Gleixner authored
Convert #MF to IDTENTRY_ERRORCODE: - Implement the C entry point with DEFINE_IDTENTRY - Emit the ASM stub with DECLARE_IDTENTRY - Remove the ASM idtentry in 64bit - Remove the open coded ASM entry code in 32bit - Fixup the XEN/PV code - Remove the old prototypes - Remove the RCU warning as the new entry macro ensures correctness No functional change. Signed-off-by:
-
