Fw: next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)

Fw: next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 3031 bytes --]

Commit

  a148866489fb ("sched: Replace rq::wake_list")

added

BUILD_BUG_ON(offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) !=
		     offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist));

in kernel/smp.c.  This seems to be failing - at least for gcc-8 (I
haven't seen any failures in my linux-next testing).

The pending-fixes branch in linux-next isjust Linus' tree plus the
branches of fixes for Linus' tree.

Begin forwarded message:

Date: Thu, 04 Jun 2020 18:15:15 -0700 (PDT)
From: "kernelci.org bot" <bot@kernelci.org>
To: linux-next@vger.kernel.org
Subject: next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)


next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)

Full Build Summary: https://kernelci.org/build/next/branch/pending-fixes/kernel/v5.7-8671-g086779e58c68/

Tree: next
Branch: pending-fixes
Git Describe: v5.7-8671-g086779e58c68
Git Commit: 086779e58c6867c3f8488d6729e267094a0cd826
Git URL: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
Built: 7 unique architectures

Build Failures Detected:

arm64:
    allmodconfig: (gcc-8) FAIL

x86_64:
    allmodconfig: (gcc-8) FAIL

Errors summary:

    1    include/linux/compiler.h:403:38: error: call to ‘__compiletime_assert_399’ declared with attribute error: BUILD_BUG_ON failed: offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) != offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist)
    1    include/linux/compiler.h:403:38: error: call to ‘__compiletime_assert_134’ declared with attribute error: BUILD_BUG_ON failed: offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) != offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist)

--------------------------------------------------------------------------------
allmodconfig (arm64, gcc-8) — FAIL, 1 error, 0 warnings, 0 section mismatches

Errors:
    include/linux/compiler.h:403:38: error: call to ‘__compiletime_assert_134’ declared with attribute error: BUILD_BUG_ON failed: offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) != offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist)

--------------------------------------------------------------------------------
allmodconfig (x86_64, gcc-8) — FAIL, 1 error, 0 warnings, 0 section mismatches

Errors:
    include/linux/compiler.h:403:38: error: call to ‘__compiletime_assert_399’ declared with attribute error: BUILD_BUG_ON failed: offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) != offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist)

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: Fw: next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)

[Peter Zijlstra]

On Fri, Jun 05, 2020 at 11:52:13AM +1000, Stephen Rothwell wrote:
> Commit
> 
>   a148866489fb ("sched: Replace rq::wake_list")
> 
> added
> 
> BUILD_BUG_ON(offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) !=
> 		     offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist));
> 
> in kernel/smp.c.  This seems to be failing - at least for gcc-8 (I
> haven't seen any failures in my linux-next testing).

CONFIG_GCC_PLUGIN_RANDSTRUCT=y

working on it.

Re: Fw: next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)

[Kees Cook]

On Fri, Jun 05, 2020 at 10:38:55AM +0200, Peter Zijlstra wrote:
> On Fri, Jun 05, 2020 at 11:52:13AM +1000, Stephen Rothwell wrote:
> > Commit
> > 
> >   a148866489fb ("sched: Replace rq::wake_list")
> > 
> > added
> > 
> > BUILD_BUG_ON(offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) !=
> > 		     offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist));
> > 
> > in kernel/smp.c.  This seems to be failing - at least for gcc-8 (I
> > haven't seen any failures in my linux-next testing).
> 
> CONFIG_GCC_PLUGIN_RANDSTRUCT=y

(FWIW, the plugins will only be enabled if the gcc plugins-dev package is
installed)

> working on it.

Either of these solutions should fix it. I'm not sure which is
preferred.


Collect them into a separate anonymous struct to keep them together
under randomization:

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 613bf7a21ae2..0f90239501c8 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -654,8 +654,11 @@ struct task_struct {
 	unsigned int			ptrace;
 
 #ifdef CONFIG_SMP
-	struct llist_node		wake_entry;
-	unsigned int			wake_entry_type;
+	/* Keep these members together under __randomize_struct. */
+	struct {
+		struct llist_node		wake_entry;
+		unsigned int			wake_entry_type;
+	};
 	int				on_cpu;
 #ifdef CONFIG_THREAD_INFO_IN_TASK
 	/* Current CPU: */


or just explicitly move them out of the randomized section:

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 613bf7a21ae2..f97b41ce1b13 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -641,6 +641,10 @@ struct task_struct {
 	/* -1 unrunnable, 0 runnable, >0 stopped: */
 	volatile long			state;
 
+#ifdef CONFIG_SMP
+	struct llist_node		wake_entry;
+	unsigned int			wake_entry_type;
+#endif
 	/*
 	 * This begins the randomizable portion of task_struct. Only
 	 * scheduling-critical items should be added above here.
@@ -654,8 +658,6 @@ struct task_struct {
 	unsigned int			ptrace;
 
 #ifdef CONFIG_SMP
-	struct llist_node		wake_entry;
-	unsigned int			wake_entry_type;
 	int				on_cpu;
 #ifdef CONFIG_THREAD_INFO_IN_TASK
 	/* Current CPU: */

-- 
Kees Cook

Re: Fw: next/pending-fixes build: 162 builds: 2 failed, 160 passed, 2 errors, 109 warnings (v5.7-8671-g086779e58c68)

[Peter Zijlstra]

On Fri, Jun 05, 2020 at 07:36:02AM -0700, Kees Cook wrote:
> On Fri, Jun 05, 2020 at 10:38:55AM +0200, Peter Zijlstra wrote:
> > On Fri, Jun 05, 2020 at 11:52:13AM +1000, Stephen Rothwell wrote:
> > > Commit
> > > 
> > >   a148866489fb ("sched: Replace rq::wake_list")
> > > 
> > > added
> > > 
> > > BUILD_BUG_ON(offsetof(struct task_struct, wake_entry_type) - offsetof(struct task_struct, wake_entry) !=
> > > 		     offsetof(struct __call_single_data, flags) - offsetof(struct __call_single_data, llist));
> > > 
> > > in kernel/smp.c.  This seems to be failing - at least for gcc-8 (I
> > > haven't seen any failures in my linux-next testing).
> > 
> > CONFIG_GCC_PLUGIN_RANDSTRUCT=y
> 
> (FWIW, the plugins will only be enabled if the gcc plugins-dev package is
> installed)

Which I hadn't, which is why I'd never seen it. I have now.

> > working on it.
> 
> Either of these solutions should fix it. I'm not sure which is
> preferred.

I posted in another thread that had earlier raised the issue:

https://lkml.kernel.org/r/20200605093704.GB2948@hirez.programming.kicks-ass.net