linux-kernel - [PATCH 3/7] proc/kcore: fix memory hotplug vs multiple opens race

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <904e8cabeddf594c30ba5fd314e2d5b887ceb3c8.1530904769.git.osandov@fb.com>
Date:   Fri,  6 Jul 2018 12:32:34 -0700
From:   Omar Sandoval <osandov@...ndov.com>
To:     linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        Alexey Dobriyan <adobriyan@...il.com>
Cc:     Eric Biederman <ebiederm@...ssion.com>, kernel-team@...com
Subject: [PATCH 3/7] proc/kcore: fix memory hotplug vs multiple opens race

From: Omar Sandoval <osandov@...com>

There's a theoretical race condition that will cause /proc/kcore to miss
a memory hotplug event:

CPU0                              CPU1
// hotplug event 1
kcore_need_update = 1

open_kcore()                      open_kcore()
    kcore_update_ram()                kcore_update_ram()
        // Walk RAM                       // Walk RAM
        __kcore_update_ram()              __kcore_update_ram()
            kcore_need_update = 0

// hotplug event 2
kcore_need_update = 1
                                              kcore_need_update = 0

Note that CPU1 set up the RAM kcore entries with the state after hotplug
event 1 but cleared the flag for hotplug event 2. The RAM entries will
therefore be stale until there is another hotplug event.

This is an extremely unlikely sequence of events, but the fix makes the
synchronization saner, anyways: we serialize the entire update sequence,
which means that whoever clears the flag will always succeed in
replacing the kcore list.

Signed-off-by: Omar Sandoval <osandov@...com>
---
 fs/proc/kcore.c | 93 +++++++++++++++++++++++--------------------------
 1 file changed, 44 insertions(+), 49 deletions(-)

diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
index eb1be07bdb3d..f335400300d3 100644
--- a/fs/proc/kcore.c
+++ b/fs/proc/kcore.c
@@ -98,53 +98,15 @@ static size_t get_kcore_size(int *nphdr, size_t *elf_buflen)
 	return size + *elf_buflen;
 }
 
-static void free_kclist_ents(struct list_head *head)
-{
-	struct kcore_list *tmp, *pos;
-
-	list_for_each_entry_safe(pos, tmp, head, list) {
-		list_del(&pos->list);
-		kfree(pos);
-	}
-}
-/*
- * Replace all KCORE_RAM/KCORE_VMEMMAP information with passed list.
- */
-static void __kcore_update_ram(struct list_head *list)
-{
-	int nphdr;
-	size_t size;
-	struct kcore_list *tmp, *pos;
-	LIST_HEAD(garbage);
-
-	down_write(&kclist_lock);
-	if (atomic_cmpxchg(&kcore_need_update, 1, 0)) {
-		list_for_each_entry_safe(pos, tmp, &kclist_head, list) {
-			if (pos->type == KCORE_RAM
-				|| pos->type == KCORE_VMEMMAP)
-				list_move(&pos->list, &garbage);
-		}
-		list_splice_tail(list, &kclist_head);
-	} else
-		list_splice(list, &garbage);
-	proc_root_kcore->size = get_kcore_size(&nphdr, &size);
-	up_write(&kclist_lock);
-
-	free_kclist_ents(&garbage);
-}
-
-
 #ifdef CONFIG_HIGHMEM
 /*
  * If no highmem, we can assume [0...max_low_pfn) continuous range of memory
  * because memory hole is not as big as !HIGHMEM case.
  * (HIGHMEM is special because part of memory is _invisible_ from the kernel.)
  */
-static int kcore_update_ram(void)
+static int kcore_ram_list(struct list_head *head)
 {
-	LIST_HEAD(head);
 	struct kcore_list *ent;
-	int ret = 0;
 
 	ent = kmalloc(sizeof(*ent), GFP_KERNEL);
 	if (!ent)
@@ -152,9 +114,8 @@ static int kcore_update_ram(void)
 	ent->addr = (unsigned long)__va(0);
 	ent->size = max_low_pfn << PAGE_SHIFT;
 	ent->type = KCORE_RAM;
-	list_add(&ent->list, &head);
-	__kcore_update_ram(&head);
-	return ret;
+	list_add(&ent->list, head);
+	return 0;
 }
 
 #else /* !CONFIG_HIGHMEM */
@@ -253,11 +214,10 @@ kclist_add_private(unsigned long pfn, unsigned long nr_pages, void *arg)
 	return 1;
 }
 
-static int kcore_update_ram(void)
+static int kcore_ram_list(struct list_head *list)
 {
 	int nid, ret;
 	unsigned long end_pfn;
-	LIST_HEAD(head);
 
 	/* Not inialized....update now */
 	/* find out "max pfn" */
@@ -269,15 +229,50 @@ static int kcore_update_ram(void)
 			end_pfn = node_end;
 	}
 	/* scan 0 to max_pfn */
-	ret = walk_system_ram_range(0, end_pfn, &head, kclist_add_private);
-	if (ret) {
-		free_kclist_ents(&head);
+	ret = walk_system_ram_range(0, end_pfn, list, kclist_add_private);
+	if (ret)
 		return -ENOMEM;
+	return 0;
+}
+#endif /* CONFIG_HIGHMEM */
+
+static int kcore_update_ram(void)
+{
+	LIST_HEAD(list);
+	LIST_HEAD(garbage);
+	int nphdr;
+	size_t size;
+	struct kcore_list *tmp, *pos;
+	int ret = 0;
+
+	down_write(&kclist_lock);
+	if (!atomic_cmpxchg(&kcore_need_update, 1, 0))
+		goto out;
+
+	ret = kcore_ram_list(&list);
+	if (ret) {
+		/* Couldn't get the RAM list, try again next time. */
+		atomic_set(&kcore_need_update, 1);
+		list_splice_tail(&list, &garbage);
+		goto out;
+	}
+
+	list_for_each_entry_safe(pos, tmp, &kclist_head, list) {
+		if (pos->type == KCORE_RAM || pos->type == KCORE_VMEMMAP)
+			list_move(&pos->list, &garbage);
+	}
+	list_splice_tail(&list, &kclist_head);
+
+	proc_root_kcore->size = get_kcore_size(&nphdr, &size);
+
+out:
+	up_write(&kclist_lock);
+	list_for_each_entry_safe(pos, tmp, &garbage, list) {
+		list_del(&pos->list);
+		kfree(pos);
 	}
-	__kcore_update_ram(&head);
 	return ret;
 }
-#endif /* CONFIG_HIGHMEM */
 
 /*****************************************************************************/
 /*
-- 
2.18.0