Tested with SATA/Opal2 and SAS/Enterprise disks.

Those features I found implemented on Samsung 970 Pro NVMe.

Previous code was not reporting errors reported by the device.

I do not completely understand how this code supposed to work,
all I see is that it just does not.

For SATA and NVMe devices we know explicitly whether they support
Trusted Computing feature set and Security Send/Receive commands.
This prevents errors from happening and being reported by kernel.

This is just for future reference.

This code handles both ATA and SCSI via CAM interface, and potentially
may handle NVMe.

That parameter is optional even for Enterprise, and we found some
devices failing StartSession when it is specified.

There may be some rough edges, but with this I successfully initialized,
locked, unlocked and reverted Intel Optane 905 SSD, implementing Pyrite 1.0.

There are drives uing both upper and lower case symbols.

Should be no functional changes.

There are some Toshiba NVMe SSDs that fail attempts to enable PBA
or load it.  Since I don't think why it should be critical for devices
not used for boot, ignore that error.  Same time leave it there just
to keep backward compatibility.

It is a part of Pyrite 2.0 SSC.

On different Enterprise drives both of them appear to be useful.

Package sedutil for debian

These two fields are uint8_t (unsigned char), causing them to be treated
as literal char instead of int values. Need to cast to some int type to fix
that.

Fix PIN dump in --query output

Fix verbose (-vvv..) command line arguments.

We've found thet some SAS/SATA interposers report incorrect error
"ILLEGAL REQUEST asc:25,0 (Logical unit not supported)" when device
behind them does not support TCG, that cause FreeBSD CAM to detach
the LUN.  To workaround that check whether SCSI device reports ATA
Information VPD page, and if so, that it reports TCG support there.

It is pretty confusing to see errors if caller is going to handle
them, either with retrial or custom reporting.  Add parameter to
mute the logging and use it in DtaSession::start() when retrying
without SessionTimeout token.

I am not sure how it worked before, may be previous Linux kernels
reported device errors as ioctl errors, but at least now it is no so.
Extend error handling to SCSI, driver and host statuses, that fixes
SAS disks operation, since otherwise due to lack of error handling
for IDENTIFY request they were incorrectly handled as SATA.

While there, alike to FreeBSD add check for TCG bit in ATA IDENTIFY
data to avoid unneeded unsupported commands errors later.

The sedutil scan currently does not display drives other than sdX.
In this commit, we extend the scan to include sdXYZ and, while we're
at it, also add the scan for nvmeXYZ drives.

sedutil uses the serial number to hash the password for NVMe disk types
and the model number for SCSI disk types. Tri-mode NVMe devices are
exposed as SCSI devices, making them incompatible on non-tri-mode HBAs.
This patch automates disk type selection by first checking if the disk
is compatible with the NVMe IOCTL interface using the NVME_IOCTL_ID
command. If it's not, it falls back to SCSI.