[PATCH] mtd: core: Fix ERROR of refcount due to del_mtd_device()

Shang XiaoJing shangxiaojing at huawei.com
Fri Nov 18 22:39:15 PST 2022


del_mtd_device() will call of_node_put() to mtd_get_of_node(mtd), which
is mtd->dev.of_node. However, memset(&mtd->dev, 0) is called before
of_node_put(). As the result, of_node_put() won't do anything in
del_mtd_device(), and causes the refcount leak.

del_mtd_device()
    memset(&mtd->dev, 0, sizeof(mtd->dev) # clear mtd->dev
    of_node_put()
        mtd_get_of_node(mtd) # mtd->dev is cleared, can't locate of_node
                             # of_node_put(NULL) won't do anything

Fix the ERROR by caching the pointer of the device_node.

OF: ERROR: memory leak, expected refcount 1 instead of 2,
of_node_get()/of_node_put() unbalanced - destroy cset entry: attach
overlay node /spi/spi-sram at 0
CPU: 3 PID: 275 Comm: python3 Tainted: G N 6.1.0-rc3+ #54
    0d8a1edddf51f172ff5226989a7565c6313b08e2
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
rel-1.15.0-0-g2dd4b9b3f840-prebuilt.qemu.org 04/01/2014
Call Trace:
<TASK>
    dump_stack_lvl+0x67/0x83
    kobject_get+0x155/0x160
    of_node_get+0x1f/0x30
    of_fwnode_get+0x43/0x70
    fwnode_handle_get+0x54/0x80
    fwnode_get_nth_parent+0xc9/0xe0
    fwnode_full_name_string+0x3f/0xa0
    device_node_string+0x30f/0x750
    pointer+0x598/0x7a0
    vsnprintf+0x62d/0x9b0
    ...
    cfs_overlay_release+0x30/0x90
    config_item_release+0xbe/0x1a0
    config_item_put+0x5e/0x80
    configfs_rmdir+0x3bd/0x540
    vfs_rmdir+0x18c/0x320
    do_rmdir+0x198/0x330
    __x64_sys_rmdir+0x2c/0x40
    do_syscall_64+0x37/0x90
    entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: 00596576a051 ("mtd: core: clear out unregistered devices a bit more")
Signed-off-by: Shang XiaoJing <shangxiaojing at huawei.com>
---
 drivers/mtd/mtdcore.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index 18aa54460d36..74b1a1c61b39 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -774,6 +774,7 @@ int del_mtd_device(struct mtd_info *mtd)
 {
 	int ret;
 	struct mtd_notifier *not;
+	struct device_node *mtd_of_node;
 
 	mutex_lock(&mtd_table_mutex);
 
@@ -792,6 +793,7 @@ int del_mtd_device(struct mtd_info *mtd)
 		       mtd->index, mtd->name, mtd->usecount);
 		ret = -EBUSY;
 	} else {
+		mtd_of_node = mtd_get_of_node(mtd);
 		debugfs_remove_recursive(mtd->dbg.dfs_dir);
 
 		/* Try to remove the NVMEM provider */
@@ -803,7 +805,7 @@ int del_mtd_device(struct mtd_info *mtd)
 		memset(&mtd->dev, 0, sizeof(mtd->dev));
 
 		idr_remove(&mtd_idr, mtd->index);
-		of_node_put(mtd_get_of_node(mtd));
+		of_node_put(mtd_of_node);
 
 		module_put(THIS_MODULE);
 		ret = 0;
-- 
2.17.1




More information about the linux-mtd mailing list