| Message ID | 1432308828-31185-1-git-send-email-Anand.Jain@oracle.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
On Fri, May 22, 2015 at 11:33:48PM +0800, Anand Jain wrote: > +void btrfs_free_stale_device(struct btrfs_device *cur_dev) > +{ > + int del = 0; > + struct btrfs_fs_devices *fs_devs; > + struct btrfs_device *dev; > + > + if (!rcu_str_deref(cur_dev->name)) > + return; This looks like rcu-unprotected access, though there's an outer mutext held in device_list_add that calls btrfs_free_stale_device. I'm not sure that the device name should be used to do any sorts of checks at all. > + list_for_each_entry(fs_devs, &fs_uuids, list) { > + if (fs_devs->opened) > + continue; > + if (fs_devs->seeding) > + continue; > + list_for_each_entry(dev, &fs_devs->devices, dev_list) { > + if (dev == cur_dev) > + continue; > + > + /* > + * Todo: This won't be enough. What if same device > + * comes back with new uuid and with its mapper path? > + * But for now, this does helps as mostly an admin will > + * use either mapper or non mapper path throughout. > + */ > + if (!rcu_str_deref(dev->name)) > + continue; > + if (!strcmp(rcu_str_deref(dev->name), > + rcu_str_deref(cur_dev->name))) { > + del = 1; > + break; > + } > + } > + if (del) { > + /* delete the stale */ > + if (fs_devs->num_devices == 1) { > + btrfs_sysfs_remove_fsid(fs_devs); > + list_del(&fs_devs->list); > + free_fs_devices(fs_devs); > + } else { > + fs_devs->num_devices--; > + list_del(&dev->dev_list); > + rcu_string_free(dev->name); > + kfree(dev); Devices are normally freed by the rcu through free_device, this looks suspicious to mix both approaches. > + } > + break; > + } > + } > + return; Unnecessary return > +} > + > /* > * Add new device to list of registered devices > * > @@ -560,6 +609,8 @@ static noinline int device_list_add(const char *path, > if (!fs_devices->opened) > device->generation = found_transid; > > + btrfs_free_stale_device(device); It might be safe to do that in the end, but it should be explained somewhere. > + > *fs_devices_ret = fs_devices; > > return ret; -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi David, Thanks for comments, more below.. On 05/27/2015 07:34 PM, David Sterba wrote: > On Fri, May 22, 2015 at 11:33:48PM +0800, Anand Jain wrote: >> +void btrfs_free_stale_device(struct btrfs_device *cur_dev) >> +{ >> + int del = 0; >> + struct btrfs_fs_devices *fs_devs; >> + struct btrfs_device *dev; >> + >> + if (!rcu_str_deref(cur_dev->name)) >> + return; > > This looks like rcu-unprotected access, though there's an outer mutext > held in device_list_add that calls btrfs_free_stale_device. I'm not sure > that the device name should be used to do any sorts of checks at all. you are right. I got this corrected in v5.2 just sent out. >> + list_for_each_entry(fs_devs, &fs_uuids, list) { >> + if (fs_devs->opened) >> + continue; >> + if (fs_devs->seeding) >> + continue; >> + list_for_each_entry(dev, &fs_devs->devices, dev_list) { >> + if (dev == cur_dev) >> + continue; >> + >> + /* >> + * Todo: This won't be enough. What if same device >> + * comes back with new uuid and with its mapper path? >> + * But for now, this does helps as mostly an admin will >> + * use either mapper or non mapper path throughout. >> + */ >> + if (!rcu_str_deref(dev->name)) >> + continue; >> + if (!strcmp(rcu_str_deref(dev->name), >> + rcu_str_deref(cur_dev->name))) { >> + del = 1; >> + break; >> + } >> + } >> + if (del) { >> + /* delete the stale */ >> + if (fs_devs->num_devices == 1) { >> + btrfs_sysfs_remove_fsid(fs_devs); >> + list_del(&fs_devs->list); >> + free_fs_devices(fs_devs); >> + } else { >> + fs_devs->num_devices--; >> + list_del(&dev->dev_list); >> + rcu_string_free(dev->name); >> + kfree(dev); > > Devices are normally freed by the rcu through free_device, this looks > suspicious to mix both approaches. yes its bit of mixed up, also in other parts as well, for eg: free_fs_devices(). Here I am following free_fs_devices, since free_stale will check on the devices that are unmounted. >> + } >> + break; >> + } >> + } >> + return; > > Unnecessary return right. >> +} >> + >> /* >> * Add new device to list of registered devices >> * >> @@ -560,6 +609,8 @@ static noinline int device_list_add(const char *path, >> if (!fs_devices->opened) >> device->generation = found_transid; >> >> + btrfs_free_stale_device(device); > > It might be safe to do that in the end, but it should be explained > somewhere. Added, Thanks, Anand >> + >> *fs_devices_ret = fs_devices; >> >> return ret; > -- > To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" 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/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 747241f..eb57331 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -445,6 +445,55 @@ static void pending_bios_fn(struct btrfs_work *work) run_scheduled_bios(device); } +void btrfs_free_stale_device(struct btrfs_device *cur_dev) +{ + int del = 0; + struct btrfs_fs_devices *fs_devs; + struct btrfs_device *dev; + + if (!rcu_str_deref(cur_dev->name)) + return; + list_for_each_entry(fs_devs, &fs_uuids, list) { + if (fs_devs->opened) + continue; + if (fs_devs->seeding) + continue; + list_for_each_entry(dev, &fs_devs->devices, dev_list) { + if (dev == cur_dev) + continue; + + /* + * Todo: This won't be enough. What if same device + * comes back with new uuid and with its mapper path? + * But for now, this does helps as mostly an admin will + * use either mapper or non mapper path throughout. + */ + if (!rcu_str_deref(dev->name)) + continue; + if (!strcmp(rcu_str_deref(dev->name), + rcu_str_deref(cur_dev->name))) { + del = 1; + break; + } + } + if (del) { + /* delete the stale */ + if (fs_devs->num_devices == 1) { + btrfs_sysfs_remove_fsid(fs_devs); + list_del(&fs_devs->list); + free_fs_devices(fs_devs); + } else { + fs_devs->num_devices--; + list_del(&dev->dev_list); + rcu_string_free(dev->name); + kfree(dev); + } + break; + } + } + return; +} + /* * Add new device to list of registered devices * @@ -560,6 +609,8 @@ static noinline int device_list_add(const char *path, if (!fs_devices->opened) device->generation = found_transid; + btrfs_free_stale_device(device); + *fs_devices_ret = fs_devices; return ret;
When the btrfs on a device is overwritten with a new btrfs (mkfs), the old btrfs instance in the kernel becomes stale. So with this patch if kernel finds device is overwritten, then delete the stale fsid/uuid. Signed-off-by: Anand Jain <anand.jain@oracle.com> --- V5->V5.1: since this deals with only devices in unmounted state, don't try to remove device link fs/btrfs/volumes.c | 51 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 51 insertions(+)