perf/core: Make sure the ring-buffer is mapped in all page-tables

The ring-buffer is accessed in the NMI handler, so it's better to avoid
faulting on it. Sync the vmalloc range with all page-tables in system to
make sure everyone has it mapped.

This fixes a WARN_ON_ONCE() that can be triggered with PTI enabled on
x86-32:

  WARNING: CPU: 4 PID: 0 at arch/x86/mm/fault.c:320 vmalloc_fault+0x220/0x230

This triggers because with PTI enabled on an PAE kernel the PMDs are no
longer shared between the page-tables, so the vmalloc changes do not
propagate automatically.

Note: Andy said rightfully that we should try to fix the vmalloc code for
that case, but that's not a hot fix for the issue at hand.

Fixes: 7757d607c6b3 ("x86/pti: Allow CONFIG_PAGE_TABLE_ISOLATION for x86_32")
Signed-off-by: Joerg Roedel <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: "H . Peter Anvin" <[email protected]>
Cc: [email protected]
Cc: Linus Torvalds <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Josh Poimboeuf <[email protected]>
Cc: Juergen Gross <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Jiri Kosina <[email protected]>
Cc: Boris Ostrovsky <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: David Laight <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: Eduardo Valentin <[email protected]>
Cc: Greg KH <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: Andrea Arcangeli <[email protected]>
Cc: Waiman Long <[email protected]>
Cc: Pavel Machek <[email protected]>
Cc: "David H . Gutteridge" <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]

1 files changed, 16 insertions, 0 deletions

diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c
index 5d3cf407e374..df2d8cf0072c 100644
--- a/kernel/events/ring_buffer.c
+++ b/kernel/events/ring_buffer.c
@@ -814,6 +814,13 @@ static void rb_free_work(struct work_struct *work)
 
 	vfree(base);
 	kfree(rb);
+
+	/*
+	 * FIXME: PAE workaround for vmalloc_fault(): Make sure buffer is
+	 * unmapped in all page-tables.
+	 */
+	if (IS_ENABLED(CONFIG_X86_PAE))
+		vmalloc_sync_all();
 }
 
 void rb_free(struct ring_buffer *rb)
@@ -840,6 +847,15 @@ struct ring_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
 	if (!all_buf)
 		goto fail_all_buf;
 
+	/*
+	 * FIXME: PAE workaround for vmalloc_fault(): The buffer is
+	 * accessed in NMI handlers, make sure it is mapped in all
+	 * page-tables in the system so that we don't fault on the range in
+	 * an NMI handler.
+	 */
+	if (IS_ENABLED(CONFIG_X86_PAE))
+		vmalloc_sync_all();
+
 	rb->user_page = all_buf;
 	rb->data_pages[0] = all_buf + PAGE_SIZE;
 	if (nr_pages) {