Message ID | 20191112035752.8338-3-bvanassche@acm.org (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
Series | Three reliability improvements | expand |
On Mon, Nov 11, 2019 at 07:57:51PM -0800, Bart Van Assche wrote: > The target_remove_from_state_list() call uses the cmd->dev pointer. Make > sure that that pointer remains valid as long as TMF processing is in > progress. This patch fixes the following KASAN complaint: > > BUG: KASAN: use-after-free in target_remove_from_state_list+0x22/0x130 [target_core_mod] > Read of size 8 at addr ffff8880b110cf80 by task kworker/0:1/12 > > CPU: 0 PID: 12 Comm: kworker/0:1 Not tainted 5.4.0-rc1-dbg+ #6 > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014 > Workqueue: events target_tmr_work [target_core_mod] > Call Trace: > dump_stack+0x8a/0xd6 > print_address_description.constprop.0+0x40/0x60 > __kasan_report.cold+0x1b/0x33 > kasan_report+0x16/0x20 > __asan_load8+0x58/0x90 > target_remove_from_state_list+0x22/0x130 [target_core_mod] > transport_cmd_check_stop_to_fabric+0x17/0xe0 [target_core_mod] > target_tmr_work+0xe2/0x1a0 [target_core_mod] > process_one_work+0x549/0xa40 > worker_thread+0x7a/0x5d0 > kthread+0x1bc/0x210 > ret_from_fork+0x24/0x30 > > The buggy address belongs to the page: > page:ffffea0002c44300 refcount:0 mapcount:0 mapping:0000000000000000 index:0x0 > flags: 0x1fff000000000000() > raw: 1fff000000000000 0000000000000000 ffffea0002c44308 0000000000000000 > raw: 0000000000000000 0000000000000000 00000000ffffffff 0000000000000000 > page dumped because: kasan: bad access detected > > Memory state around the buggy address: > ffff8880b110ce80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff > ffff8880b110cf00: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff > >ffff8880b110cf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff > ^ > ffff8880b110d000: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff > ffff8880b110d080: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff > > Cc: Mike Christie <mchristi@redhat.com> > Cc: Roman Bolshakov <r.bolshakov@yadro.com> > Signed-off-by: Bart Van Assche <bvanassche@acm.org> > --- > drivers/target/target_core_transport.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c > index 09f833c1de8a..ba6bd73fe63b 100644 > --- a/drivers/target/target_core_transport.c > +++ b/drivers/target/target_core_transport.c > @@ -3256,6 +3256,8 @@ static void target_tmr_work(struct work_struct *work) > struct se_tmr_req *tmr = cmd->se_tmr_req; > int ret; > > + target_depend_item(&dev->dev_group.cg_item); > + > if (cmd->transport_state & CMD_T_ABORTED) > goto aborted; > > @@ -3297,10 +3299,14 @@ static void target_tmr_work(struct work_struct *work) > cmd->se_tfo->queue_tm_rsp(cmd); > > transport_cmd_check_stop_to_fabric(cmd); > + > +out: > + target_undepend_item(&dev->dev_group.cg_item); > return; > > aborted: > target_handle_abort(cmd); > + goto out; > } > > int transport_generic_handle_tmr( > -- > 2.23.0 > Hi Bart, As far as I understand the change prevents backstore directory removal from TCM configfs when a TMF is processed and targetcli would start to fail sporadically as it tries to delete configfs directories only once: https://github.com/open-iscsi/rtslib-fb/blob/master/rtslib/node.py#L228 Should we hold se_device device without preventing removal of backstore? Thanks, Roman
On 11/13/19 4:02 AM, Roman Bolshakov wrote: > As far as I understand the change prevents backstore directory removal > from TCM configfs when a TMF is processed and targetcli would start > to fail sporadically as it tries to delete configfs directories only once: > https://github.com/open-iscsi/rtslib-fb/blob/master/rtslib/node.py#L228 > > Should we hold se_device device without preventing removal of backstore? How about using config_item_get()/config_item_put() instead of target_depend_item()/target_undepend_item()? Thanks, Bart.
On Wed, Nov 13, 2019 at 09:05:15AM -0800, Bart Van Assche wrote: > On 11/13/19 4:02 AM, Roman Bolshakov wrote: > > As far as I understand the change prevents backstore directory removal > > from TCM configfs when a TMF is processed and targetcli would start > > to fail sporadically as it tries to delete configfs directories only once: > > https://github.com/open-iscsi/rtslib-fb/blob/master/rtslib/node.py#L228 > > > > Should we hold se_device device without preventing removal of backstore? > > How about using config_item_get()/config_item_put() instead of > target_depend_item()/target_undepend_item()? > > Thanks, > > Bart. Yes, config_item_get()/config_item_put() or config_group_get()/config_group_put() seem to be better solutions. Thanks, Roman
diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c index 09f833c1de8a..ba6bd73fe63b 100644 --- a/drivers/target/target_core_transport.c +++ b/drivers/target/target_core_transport.c @@ -3256,6 +3256,8 @@ static void target_tmr_work(struct work_struct *work) struct se_tmr_req *tmr = cmd->se_tmr_req; int ret; + target_depend_item(&dev->dev_group.cg_item); + if (cmd->transport_state & CMD_T_ABORTED) goto aborted; @@ -3297,10 +3299,14 @@ static void target_tmr_work(struct work_struct *work) cmd->se_tfo->queue_tm_rsp(cmd); transport_cmd_check_stop_to_fabric(cmd); + +out: + target_undepend_item(&dev->dev_group.cg_item); return; aborted: target_handle_abort(cmd); + goto out; } int transport_generic_handle_tmr(