mbox series

[0/7] usb-storage,uas: Support OPAL commands on USB attached devices.

Message ID 20231016072604.40179-1-gmazyland@gmail.com (mailing list archive)
Headers show
Series usb-storage,uas: Support OPAL commands on USB attached devices. | expand

Message

Milan Broz Oct. 16, 2023, 7:25 a.m. UTC
This patchset adds support for OPAL commands (self-encrypted drives)
through USB-attached storage (usb-storage and UAS drivers).

The related SCSI change was sent in a separate patch
https://lore.kernel.org/linux-scsi/20231016070211.39502-1-gmazyland@gmail.com/

The first part (64-bit quirks) is generic and will be needed later anyway
once new flags appear.

1) Patches 1-5 only add support for 64-bit quirks for USB storage
(unfortunately, USB device info can be 32-bit on 32-bit platforms,
and we are out of space for flags now).

2) Patches 6-7 enable OPAL commands on USB devices and adds
IGNORE_OPAL quirk. The last patch uses the flag for Realtek 9210
devices that do not behave correctly.

More info

1) 64bit USB storage quirk flags

The quirks are transferred through the device info value, which
is unsigned long (and as a part of USB infrastructure, it cannot
be changed).
After discussion on the USB list, I used high bit as an indicator
that the values need to be translated/unpacked to 64bit
(while lower values are used directly).

This is implemented through a host-compiled program that
generates device tables and uses a translation function.
As both usb-storage and UAS drivers share a lot of headers and
definitions, we need to generate separate files for usb-storage
and UAS. Note that due to the linking of both UAS and mass-storage
together, it must use separate names for translation tables.

(I also tried to use a statically generated array for flags,
but this increased the size of drivers significantly and
the code was quite ugly...)

2) Support for OPAL on USB attached storage.

The main support for OPAL on USB-attached storage is
straightforward (it depends on ATA-12 pass-thru support
for security commands).
Patch 6 enables the SCSI security flag for USB mass storage
and UAS device by default.

During device detection, the USB driver (mass-storage, UAS) enables
the security driver flag to allow SED OPAL code to run OPAL discovery
command for the device. If it receives a correct reply, OPAL is enabled
for the device. If not (or if SCSI command is rejected), OPAL
remains disabled.

Enabling OPAL support may uncover many issues, as OPAL-locked devices
often tend to generate errors on the locked range.

Anyway, cryptsetup will soon support OPAL devices, and I think support
for USB devices is a nice feature that enables users to unlock drives
even if they are attached through USB adapters.

There are also bugs in firmware implementations, so I added a quirk
flag that can disable security commands for particular devices.

The last patch uses this quirk for Realtek 9210, which seems to support
OPAL commands, but after configuring OPAL locking range, it also sets
the write-protected flag for the whole device.
This is perhaps a bug in firmware (all versions I tried), and I will
report that later to Realtek.


Milan Broz (7):
  usb-storage: remove UNUSUAL_VENDOR_INTF macro
  usb-storage,uas: make internal quirks flags 64bit
  usb-storage: use fflags index only in usb-storage driver
  usb-storage,uas: use host helper to generate driver info
  usb-storage,uas: do not convert device_info for 64-bit platforms
  usb-storage,uas: enable security commands for USB-attached storage
  usb-storage,uas: disable security commands (OPAL) for RT9210 chip
    family

 .../admin-guide/kernel-parameters.txt         |   2 +
 drivers/usb/storage/Makefile                  |  28 +++
 drivers/usb/storage/alauda.c                  |   2 +-
 drivers/usb/storage/cypress_atacb.c           |   2 +-
 drivers/usb/storage/datafab.c                 |   2 +-
 drivers/usb/storage/ene_ub6250.c              |   2 +-
 drivers/usb/storage/freecom.c                 |   2 +-
 drivers/usb/storage/isd200.c                  |   2 +-
 drivers/usb/storage/jumpshot.c                |   2 +-
 drivers/usb/storage/karma.c                   |   2 +-
 drivers/usb/storage/mkflags.c                 | 235 ++++++++++++++++++
 drivers/usb/storage/onetouch.c                |   2 +-
 drivers/usb/storage/realtek_cr.c              |   2 +-
 drivers/usb/storage/scsiglue.c                |   4 +
 drivers/usb/storage/sddr09.c                  |   2 +-
 drivers/usb/storage/sddr55.c                  |   2 +-
 drivers/usb/storage/shuttle_usbat.c           |   2 +-
 drivers/usb/storage/uas-detect.h              |   6 +-
 drivers/usb/storage/uas.c                     |  29 +--
 drivers/usb/storage/unusual_devs.h            |  11 +
 drivers/usb/storage/unusual_uas.h             |  11 +
 drivers/usb/storage/usb-ids.h                 |  37 +++
 drivers/usb/storage/usb.c                     |  44 ++--
 drivers/usb/storage/usb.h                     |   7 +-
 drivers/usb/storage/usual-tables.c            |  38 +--
 include/linux/usb_usual.h                     |   2 +
 26 files changed, 385 insertions(+), 95 deletions(-)
 create mode 100644 drivers/usb/storage/mkflags.c
 create mode 100644 drivers/usb/storage/usb-ids.h

Comments

Alan Stern Oct. 16, 2023, 5:33 p.m. UTC | #1
On Mon, Oct 16, 2023 at 09:25:57AM +0200, Milan Broz wrote:
> This patchset adds support for OPAL commands (self-encrypted drives)
> through USB-attached storage (usb-storage and UAS drivers).

This is version 2 of the proposed patch set, but you didn't include the 
version number in the email Subject: lines and you didn't include the 
summary of differences from v1 below the "---" lines of the various 
patches.

Patches 5, 6, and 7 look okay.  You can add my Reviewed-by: to each of 
them.

I've got some additional comments on patch 4 (in a separate email).

Alan Stern
Milan Broz Oct. 16, 2023, 5:48 p.m. UTC | #2
On 10/16/23 19:33, Alan Stern wrote:
> On Mon, Oct 16, 2023 at 09:25:57AM +0200, Milan Broz wrote:
>> This patchset adds support for OPAL commands (self-encrypted drives)
>> through USB-attached storage (usb-storage and UAS drivers).
> 
> This is version 2 of the proposed patch set, but you didn't include the
> version number in the email Subject: lines and you didn't include the
> summary of differences from v1 below the "---" lines of the various
> patches.

Hi,

well, the first patchset was RFC, so I sent is as "the first real version".
Perhaps not the correct way, sorry for that.

Anyway, if you see the discussion about OPAL change on SCSI list, another
solution (inside USB storage driver) is needed.

So, please ignore patch 6/7, these will be needed, but I have to rewrite
SCSI logic to USB glue/UAS driver.

But for the generic 64-bit flags (patch 1-5), if you see this useful, please review it.

Common requirement is that kernel patch need an user for merge
(and my flag is currently no going to be used without rewrite).

But that time will come one day, and if I can save people time to reinvent
the 64-bit quirks logic, it would be nice to merge it.

Thanks,
Milan



> 
> Patches 5, 6, and 7 look okay.  You can add my Reviewed-by: to each of
> them.
> 
> I've got some additional comments on patch 4 (in a separate email).
> 
> Alan Stern