| Message ID | 1533286080-13099-1-git-send-email-mlombard@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RESEND,v2] target: move the rx hdr buffer out of the stack | expand |
On 08/03/2018 03:48 AM, Maurizio Lombardi wrote: > When HeaderDigest is enabled on aarch64 machines, target triggers > a lot of failed crc checksum errors: > > example of output in dmesg: > [ 154.495031] HeaderDigest CRC32C failed, received 0x60571c8d, computed 0x288c3ab9 > [ 154.583821] Got unknown iSCSI OpCode: 0xff > [ 162.979857] HeaderDigest CRC32C failed, received 0x0712e57c, computed 0x288c3ab9 > ... > > The problem is that the iscsit_get_rx_pdu() function uses > a stack buffer as input for the scatterlist crypto library. > This should be avoided on kernels >= 4.9 because stack buffers > may not be directly mappable to struct page when the > CONFIG_VMAP_STACK option is enabled. > > This patch modifies the code so the buffer will be allocated on the heap > and adds a pointer to it in the iscsi_conn structure. > > v2: allocate conn_rx_buf in iscsi_target_login_thread() > and fix a memory leak > > Signed-off-by: Maurizio Lombardi <mlombard@redhat.com> > --- > drivers/target/iscsi/iscsi_target.c | 6 +++++- > drivers/target/iscsi/iscsi_target_login.c | 15 +++++++++++++++ > include/target/iscsi/iscsi_target_core.h | 2 ++ > 3 files changed, 22 insertions(+), 1 deletion(-) > > diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c > index 8e22379..149fa91 100644 > --- a/drivers/target/iscsi/iscsi_target.c > +++ b/drivers/target/iscsi/iscsi_target.c > @@ -3913,10 +3913,12 @@ static bool iscsi_target_check_conn_state(struct iscsi_conn *conn) > static void iscsit_get_rx_pdu(struct iscsi_conn *conn) > { > int ret; > - u8 buffer[ISCSI_HDR_LEN], opcode; > + u8 *buffer, opcode; > u32 checksum = 0, digest = 0; > struct kvec iov; > > + buffer = conn->conn_rx_buf; If the buffer is only used in the loop below why does it need to be allocated in __iscsi_target_login_thread? It just seems like the error handling and freeing of that buffer would be more simple if done here. -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c index 8e22379..149fa91 100644 --- a/drivers/target/iscsi/iscsi_target.c +++ b/drivers/target/iscsi/iscsi_target.c @@ -3913,10 +3913,12 @@ static bool iscsi_target_check_conn_state(struct iscsi_conn *conn) static void iscsit_get_rx_pdu(struct iscsi_conn *conn) { int ret; - u8 buffer[ISCSI_HDR_LEN], opcode; + u8 *buffer, opcode; u32 checksum = 0, digest = 0; struct kvec iov; + buffer = conn->conn_rx_buf; + while (!kthread_should_stop()) { /* * Ensure that both TX and RX per connection kthreads @@ -4211,6 +4213,8 @@ int iscsit_close_connection( crypto_free_ahash(tfm); } + kfree(conn->conn_rx_buf); + free_cpumask_var(conn->conn_cpumask); kfree(conn->conn_ops); diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c index 9950178..4e8974a5 100644 --- a/drivers/target/iscsi/iscsi_target_login.c +++ b/drivers/target/iscsi/iscsi_target_login.c @@ -1203,6 +1203,8 @@ void iscsi_target_login_sess_out(struct iscsi_conn *conn, crypto_free_ahash(tfm); } + kfree(conn->conn_rx_buf); + free_cpumask_var(conn->conn_cpumask); kfree(conn->conn_ops); @@ -1262,6 +1264,15 @@ static int __iscsi_target_login_thread(struct iscsi_np *np) /* Get another socket */ return 1; } + + /* Allocate buffer for iscsit_get_rx_pdu() */ + conn->conn_rx_buf = kmalloc(ISCSI_HDR_LEN, GFP_KERNEL); + if (!conn->conn_rx_buf) { + pr_err("Could not allocate memory for conn_rx_buf\n"); + kfree(conn); + return 1; + } + pr_debug("Moving to TARG_CONN_STATE_FREE.\n"); conn->conn_state = TARG_CONN_STATE_FREE; @@ -1270,6 +1281,7 @@ static int __iscsi_target_login_thread(struct iscsi_np *np) timer_setup(&conn->nopin_timer, iscsit_handle_nopin_timeout, 0); if (iscsit_conn_set_transport(conn, np->np_transport) < 0) { + kfree(conn->conn_rx_buf); kfree(conn); return 1; } @@ -1278,6 +1290,7 @@ static int __iscsi_target_login_thread(struct iscsi_np *np) if (rc == -ENOSYS) { complete(&np->np_restart_comp); iscsit_put_transport(conn->conn_transport); + kfree(conn->conn_rx_buf); kfree(conn); conn = NULL; goto exit; @@ -1288,6 +1301,7 @@ static int __iscsi_target_login_thread(struct iscsi_np *np) spin_unlock_bh(&np->np_thread_lock); complete(&np->np_restart_comp); iscsit_put_transport(conn->conn_transport); + kfree(conn->conn_rx_buf); kfree(conn); conn = NULL; /* Get another socket */ @@ -1295,6 +1309,7 @@ static int __iscsi_target_login_thread(struct iscsi_np *np) } spin_unlock_bh(&np->np_thread_lock); iscsit_put_transport(conn->conn_transport); + kfree(conn->conn_rx_buf); kfree(conn); conn = NULL; goto out; diff --git a/include/target/iscsi/iscsi_target_core.h b/include/target/iscsi/iscsi_target_core.h index cf5f3ff..3e9c525 100644 --- a/include/target/iscsi/iscsi_target_core.h +++ b/include/target/iscsi/iscsi_target_core.h @@ -584,6 +584,8 @@ struct iscsi_conn { /* libcrypto RX and TX contexts for crc32c */ struct ahash_request *conn_rx_hash; struct ahash_request *conn_tx_hash; + /* Used as buffer for iscsit_get_rx_pdu */ + u8 *conn_rx_buf; /* Used for scheduling TX and RX connection kthreads */ cpumask_var_t conn_cpumask; unsigned int conn_rx_reset_cpumask:1;
When HeaderDigest is enabled on aarch64 machines, target triggers a lot of failed crc checksum errors: example of output in dmesg: [ 154.495031] HeaderDigest CRC32C failed, received 0x60571c8d, computed 0x288c3ab9 [ 154.583821] Got unknown iSCSI OpCode: 0xff [ 162.979857] HeaderDigest CRC32C failed, received 0x0712e57c, computed 0x288c3ab9 ... The problem is that the iscsit_get_rx_pdu() function uses a stack buffer as input for the scatterlist crypto library. This should be avoided on kernels >= 4.9 because stack buffers may not be directly mappable to struct page when the CONFIG_VMAP_STACK option is enabled. This patch modifies the code so the buffer will be allocated on the heap and adds a pointer to it in the iscsi_conn structure. v2: allocate conn_rx_buf in iscsi_target_login_thread() and fix a memory leak Signed-off-by: Maurizio Lombardi <mlombard@redhat.com> --- drivers/target/iscsi/iscsi_target.c | 6 +++++- drivers/target/iscsi/iscsi_target_login.c | 15 +++++++++++++++ include/target/iscsi/iscsi_target_core.h | 2 ++ 3 files changed, 22 insertions(+), 1 deletion(-)