commit acb383f1dcb4f1e79b66d4be3a0b6f519a957b0d upstream.
Richard observed a forever loop of erofs_read_raw_page() [1] which can be generated by forcely setting ->u.i_blkaddr to 0xdeadbeef (as my understanding block layer can handle access beyond end of device correctly).
After digging into that, it seems the problem is highly related with directories and then I found the root cause is an improper error handling in erofs_readdir().
Let's fix it now.
[1] https://lore.kernel.org/r/1163995781.68824.1566084358245.JavaMail.zimbra@nod...
Reported-by: Richard Weinberger richard@nod.at Fixes: 3aa8ec716e52 ("staging: erofs: add directory operations") Cc: stable@vger.kernel.org # 4.19+ Reviewed-by: Chao Yu yuchao0@huawei.com Signed-off-by: Gao Xiang gaoxiang25@huawei.com Link: https://lore.kernel.org/r/20190818125457.25906-1-hsiangkao@aol.com Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org [ Gao Xiang: Since earlier kernels don't define EFSCORRUPTED, let's use original error code instead. ] Signed-off-by: Gao Xiang gaoxiang25@huawei.com --- drivers/staging/erofs/dir.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/drivers/staging/erofs/dir.c b/drivers/staging/erofs/dir.c index dbf6a151886c..b11cecd0a21d 100644 --- a/drivers/staging/erofs/dir.c +++ b/drivers/staging/erofs/dir.c @@ -99,8 +99,15 @@ static int erofs_readdir(struct file *f, struct dir_context *ctx) unsigned int nameoff, maxsize;
dentry_page = read_mapping_page(mapping, i, NULL); - if (IS_ERR(dentry_page)) - continue; + if (dentry_page == ERR_PTR(-ENOMEM)) { + err = -ENOMEM; + break; + } else if (IS_ERR(dentry_page)) { + errln("fail to readdir of logical block %u of nid %llu", + i, EROFS_V(dir)->nid); + err = PTR_ERR(dentry_page); + break; + }
de = (struct erofs_dirent *)kmap(dentry_page);