[PATCH] eCryptfs: Quota check incorrectly ignored

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

eCryptfs recently modified the write path to perform encryption
and write down in ecryptfs_writepage(). This function invokes vfs_write()
to write down the encrypted data to lower page cache. vfs_write() will
first make sure this write will not exceed the quota limit for the owner
of the file being written into, if quota is supported by 
the underlying file system, and it is turned on. Normally, it accomplishs
this job by calling check_idq()/check_bdq() (fs/quota/dquot.c). When system
dirty ratio is not high, ecryptfs_writepage() is normally invoked by the 
write back kernel thread, who has the capability CAP_SYS_RESOURCE, 
this priority will let check_idq()/check_bdq() directly bypass the quota
check. This sometimes makes data safely written into the disk in spite of
exceeding the quota limit.

This patch temporarily removes the CAP_SYS_RESOURCE capability from the kernel
thread before invoking vfs_write_lower(), to let it undergo quota check by
the lower file system, if necessary. After that, reassign the capability.

Signed-off-by: Li Wang <liwang@xxxxxxxxxxx>
Singed-off-by: Yong Peng <pengyong@xxxxxxxxxxxxxx>

To repeat this bug, 
	mount -o usrquota /dev/sda3 /tmp
	cd /tmp
	edquota -u foo // set the disk quota limit for user foo be m1 bytes
	quotaon -a
	mount -t ecryptfs cipher plain

	login the system as user foo
	cd /tmp/plain
	execute the following simple program
	int main()
		char buf[m2]; // m2>m1
		FILE *f = fopen("dummy", "w");
		fwrite(buf, 1, m2, f);
		sleep(60); // let the kernel thread do the write back job
		return 0;
	This program can write as much of data as it wants, provided sleep enough long time before closing the file.

 fs/ecryptfs/crypto.c |   27 +++++++++++++++++++++++++++
 1 files changed, 27 insertions(+), 0 deletions(-)

diff --git a/fs/ecryptfs/crypto.c b/fs/ecryptfs/crypto.c
index 63ab245..2c1da29 100644
--- a/fs/ecryptfs/crypto.c
+++ b/fs/ecryptfs/crypto.c
@@ -457,6 +457,7 @@ int ecryptfs_encrypt_page(struct page *page)
 	struct page *enc_extent_page = NULL;
 	loff_t extent_offset;
 	int rc = 0;
+	struct cred *cred;
 	ecryptfs_inode = page->mapping->host;
 	crypt_stat =
@@ -487,8 +488,34 @@ int ecryptfs_encrypt_page(struct page *page)
 				   * (PAGE_CACHE_SIZE
 				      / crypt_stat->extent_size))
 				  + extent_offset), crypt_stat);
+		if (current->flags & PF_KTHREAD) {
+			/*
+                         * Temporarily remove the CAP_SYS_RESOURCE capability
+                         * from the write back kernel thread to let it undergo
+                         * quota check by the lower file system
+                         */
+			cred = prepare_creds();
+			if (unlikely(!cred)) {
+				rc = -ENOMEM;
+				goto out;
+			}
+			cap_lower(cred->cap_effective, CAP_SYS_RESOURCE);
+			commit_creds(cred);
+		}
 		rc = ecryptfs_write_lower(ecryptfs_inode, enc_extent_virt,
 					  offset, crypt_stat->extent_size);
+		if (current->flags & PF_KTHREAD) {
+			/*
+                         * Reassign the CAP_SYS_RESOURCE capability
+                         */
+			cred = prepare_creds();
+			if (unlikely(!cred)) {
+				rc = -ENOMEM;
+				goto out;
+			}
+			cap_raise(cred->cap_effective, CAP_SYS_RESOURCE);
+			commit_creds(cred);
+		}
 		if (rc < 0) {
 			ecryptfs_printk(KERN_ERR, "Error attempting "
 					"to write lower page; rc = [%d]"
To unsubscribe from this list: send the line "unsubscribe ecryptfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[LARTC]     [Bugtraq]     [Yosemite Forum]     [Photo]

Powered by Linux