mm: fix unexpected changes to {failslab|fail_page_alloc}.attr

When we specify __GFP_NOWARN, we only expect that no warnings will be
issued for current caller.  But in the __should_failslab() and
__should_fail_alloc_page(), the local GFP flags alter the global
{failslab|fail_page_alloc}.attr, which is persistent and shared by all
tasks.  This is not what we expected, let's fix it.

[akpm@linux-foundation.org: unexport should_fail_ex()]
Link: https://lkml.kernel.org/r/20221118100011.2634-1-zhengqi.arch@bytedance.com
Fixes: 3f913fc5f974 ("mm: fix missing handler for __GFP_NOWARN")
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Akinobu Mita <akinobu.mita@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Cc: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

diff --git a/include/linux/fault-inject.h b/include/linux/fault-inject.h
index 9f6e25467844ae1ac708a955cebbff254abf0f3c..444236dadcf0853d525d8fe99af7da2a73ad6640 100644 (file)
--- a/include/linux/fault-inject.h
+++ b/include/linux/fault-inject.h
@@ -20,7 +20,6 @@ struct fault_attr {
        atomic_t space;
        unsigned long verbose;
        bool task_filter;
-       bool no_warn;
        unsigned long stacktrace_depth;
        unsigned long require_start;
        unsigned long require_end;
@@ -32,6 +31,10 @@ struct fault_attr {
        struct dentry *dname;
 };
 
+enum fault_flags {
+       FAULT_NOWARN =  1 << 0,
+};
+
 #define FAULT_ATTR_INITIALIZER {                                       \
                .interval = 1,                                          \
                .times = ATOMIC_INIT(1),                                \
@@ -40,11 +43,11 @@ struct fault_attr {
                .ratelimit_state = RATELIMIT_STATE_INIT_DISABLED,       \
                .verbose = 2,                                           \
                .dname = NULL,                                          \
-               .no_warn = false,                                       \
        }
 
 #define DECLARE_FAULT_ATTR(name) struct fault_attr name = FAULT_ATTR_INITIALIZER
 int setup_fault_attr(struct fault_attr *attr, char *str);
+bool should_fail_ex(struct fault_attr *attr, ssize_t size, int flags);
 bool should_fail(struct fault_attr *attr, ssize_t size);
 
 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS

diff --git a/lib/fault-inject.c b/lib/fault-inject.c
index 96e092de5b72364205085aedba1e30909a3046d6..adb2f9355ee621ed68a84f0c0bd018501e399305 100644 (file)
--- a/lib/fault-inject.c
+++ b/lib/fault-inject.c
@@ -41,9 +41,6 @@ EXPORT_SYMBOL_GPL(setup_fault_attr);
 
 static void fail_dump(struct fault_attr *attr)
 {
-       if (attr->no_warn)
-               return;
-
        if (attr->verbose > 0 && __ratelimit(&attr->ratelimit_state)) {
                printk(KERN_NOTICE "FAULT_INJECTION: forcing a failure.\n"
                       "name %pd, interval %lu, probability %lu, "
@@ -103,7 +100,7 @@ static inline bool fail_stacktrace(struct fault_attr *attr)
  * http://www.nongnu.org/failmalloc/
  */
 
-bool should_fail(struct fault_attr *attr, ssize_t size)
+bool should_fail_ex(struct fault_attr *attr, ssize_t size, int flags)
 {
        if (in_task()) {
                unsigned int fail_nth = READ_ONCE(current->fail_nth);
@@ -146,13 +143,19 @@ bool should_fail(struct fault_attr *attr, ssize_t size)
                return false;
 
 fail:
-       fail_dump(attr);
+       if (!(flags & FAULT_NOWARN))
+               fail_dump(attr);
 
        if (atomic_read(&attr->times) != -1)
                atomic_dec_not_zero(&attr->times);
 
        return true;
 }
+
+bool should_fail(struct fault_attr *attr, ssize_t size)
+{
+       return should_fail_ex(attr, size, 0);
+}
 EXPORT_SYMBOL_GPL(should_fail);
 
 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS

diff --git a/mm/failslab.c b/mm/failslab.c
index 58df9789f1d22627f569e92d527c50068f276062..ffc420c0e767f097380e587cb833eb495a484d49 100644 (file)
--- a/mm/failslab.c
+++ b/mm/failslab.c
@@ -16,6 +16,8 @@ static struct {
 
 bool __should_failslab(struct kmem_cache *s, gfp_t gfpflags)
 {
+       int flags = 0;
+
        /* No fault-injection for bootstrap cache */
        if (unlikely(s == kmem_cache))
                return false;
@@ -30,10 +32,16 @@ bool __should_failslab(struct kmem_cache *s, gfp_t gfpflags)
        if (failslab.cache_filter && !(s->flags & SLAB_FAILSLAB))
                return false;
 
+       /*
+        * In some cases, it expects to specify __GFP_NOWARN
+        * to avoid printing any information(not just a warning),
+        * thus avoiding deadlocks. See commit 6b9dbedbe349 for
+        * details.
+        */
        if (gfpflags & __GFP_NOWARN)
-               failslab.attr.no_warn = true;
+               flags |= FAULT_NOWARN;
 
-       return should_fail(&failslab.attr, s->object_size);
+       return should_fail_ex(&failslab.attr, s->object_size, flags);
 }
 
 static int __init setup_failslab(char *str)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 218b28ee49ed378523c1d252d4aed0b4d5aedc05..6e60657875d3289defa20e70c116593f0ca38829 100644 (file)
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3887,6 +3887,8 @@ __setup("fail_page_alloc=", setup_fail_page_alloc);
 
 static bool __should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
 {
+       int flags = 0;
+
        if (order < fail_page_alloc.min_order)
                return false;
        if (gfp_mask & __GFP_NOFAIL)
@@ -3897,10 +3899,11 @@ static bool __should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
                        (gfp_mask & __GFP_DIRECT_RECLAIM))
                return false;
 
+       /* See comment in __should_failslab() */
        if (gfp_mask & __GFP_NOWARN)
-               fail_page_alloc.attr.no_warn = true;
+               flags |= FAULT_NOWARN;
 
-       return should_fail(&fail_page_alloc.attr, 1 << order);
+       return should_fail_ex(&fail_page_alloc.attr, 1 << order, flags);
 }
 
 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS