Replace the device number bitmap with IDR. Extend the number of devices we
can create to 64k.
Since an IDR allows us to associate a pointer with an ID, we use this now
to rewrite tpm_chip_find_get() to simply look up the chip pointer by the
given device ID.

Protect the IDR calls with a mutex.
Signed-off-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>

tpm_chip_alloc becomes a typical subsystem allocate call.
Signed-off-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Tested-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>

Now that the tpm core has strong locking around 'ops' it is possible
to remove a TPM driver, module and all, even while user space still
has things like /dev/tpmX open. For consistency and simplicity, drop
the module locking entirely.
Signed-off-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>

Add a read/write semaphore around the ops function pointers so
ops can be set to null when the driver un-registers.

Previously the tpm core expected module locking to be enough to
ensure that tpm_unregister could not be called during certain times,
however that hasn't been sufficient for a long time.

Introduce a read/write semaphore around 'ops' so the core can set
it to null when unregistering. This provides a strong fence around
the driver callbacks, guaranteeing to the driver that no callbacks
are running or will run again.

For now the ops_lock is placed very high in the call stack, it could
be pushed down and made more granular in future if necessary.
Signed-off-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>

Now that we have a proper struct device just use dev_name() to
access this value instead of keeping two copies.
Signed-off-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>

This is a hold over from before the struct device conversion.

- All prints should be using &chip->dev, which is the Linux
  standard. This changes prints to use tpm0 as the device name,
  not the PnP/etc ID.
- The few places involving sysfs/modules that really do need the
  parent just use chip->dev.parent instead
- We no longer need to get_device(pdev) in any places since it is no
  longer used by any of the code. The kref on the parent is held
  by the device core during device_add and dropped in device_del
Signed-off-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Tested-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>

Call put_device() and return error code if devm_add_action() fails.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reported-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Fixes: 8e0ee3c9 ("tpm: fix the cleanup of struct tpm_chip")

If the initialization fails before tpm_chip_register(), put_device()
will be not called, which causes release callback not to be called.
This patch fixes the issue by adding put_device() to devres list of
the parent device.

Fixes: 313d21ee ("tpm: device class for tpm")
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
cc: stable@vger.kernel.org
Reviewed-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>

Fixed the rollback and gave more self-documenting names for the
functions.

Fixes: d972b052 ("tpm: fix call order in tpm-chip.c")
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
cc: stable@vger.kernel.org
Reviewed-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>

__compat_only_sysfs_link_entry_to_kobj() was unconditionally called for
TPM1 chips, which caused crash on Acer C720 laptop where DSM for the
ACPI object did not exist.

There are two reasons for unwanted behavior:

* The code did not check whether
  __compat_only_sysfs_link_entry_to_kobj() returned -ENOENT. This is
  OK. It just meanst that ppi is not available.
* The code did not clean up properly. Compat link should added only
  after all other init is done.

This patch sorts out these issues.

Fixes: 9b774d5cReported-by: NJeremiah Mahler <jmmahler@gmail.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NJeremiah Mahler <jmmahler@gmail.com>
Acked-by: NPeter Huewe <PeterHuewe@gmx.de>

Mimi reported that afb5abc2 reverts the fix in 398a1e71. This patch
reverts it back.

Fixes: afb5abc2 ("tpm: two-phase chip management functions")
Reported-by: NMimi Zohar <zohar@linux.vnet.ibm.com>
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Acked-by: NPeter Huewe <PeterHuewe@gmx.de>

Moved PPI attributes to the character device directory. This aligns with
the sysfs guidelines and makes them race free because they are created
atomically with the character device as part of device_register().The
character device and the sysfs attributes appear at the same time to the
user space.

As part of this change we enable PPI attributes also for TPM 2.0
devices. In order to retain backwards compatibility with TPM 1.x
devices, a symlink is created to the platform device directory.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NJason Gunthorpe <jason.gunthorpe@obsidianresearch.com>
Tested-by: Mimi Zohar <zohar@linux.vnet.ibm.com> (on TPM 1.2)
Tested-by: NChris J Arges <chris.j.arges@canonical.com>
Tested-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

When a cdev is contained in a dynamic structure the cdev parent kobj
should be set to the kobj that controls the lifetime of the enclosing
structure. In TPM's case this is the embedded struct device.

Also, cdev_init 0's the whole structure, so all sets must be after,
not before. This fixes module ref counting and cdev.

Cc: <stable@vger.kernel.org>
Fixes: 313d21ee ("tpm: device class for tpm")
Signed-off-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Reviewed-by: NDmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

I started to work with PPI interface so that it would be available
under character device sysfs directory and realized that chip
registeration was still too messy.

In TPM 1.x in some rare scenarios (errors that almost never occur)
wrong order in deinitialization steps was taken in teardown. I
reproduced these scenarios by manually inserting error codes in the
place of the corresponding function calls.

The key problem is that the teardown is messy with two separate code
paths (this was inherited when moving code from tpm-interface.c).

Moved TPM 1.x specific register/unregister functionality to own helper
functions and added single code path for teardown in tpm_chip_register().
Now the code paths have been fixed and it should be easier to review
later on this part of the code.

Cc: <stable@vger.kernel.org>
Fixes: 7a1d7e6d ("tpm: TPM 2.0 baseline support")
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Tested-by: NScot Doyle <lkml14@scotdoyle.com>
Reviewed-by: NPeter Huewe <peterhuewe@gmx.de>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

- tpm_dev_add_device(): cdev_add() must be done before uevent is
  propagated in order to avoid races.
- tpm_chip_register(): tpm_dev_add_device() must be done as the
  last step before exposing device to the user space in order to
  avoid races.

In addition clarified description in tpm_chip_register().

Fixes: 313d21ee ("tpm: device class for tpm")
Fixes: afb5abc2 ("tpm: two-phase chip management functions")
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NPeter Huewe <peterhuewe@gmx.de>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

- tpm_dev_add_device(): cdev_add() must be done before uevent is
  propagated in order to avoid races.
- tpm_chip_register(): tpm_dev_add_device() must be done as the
  last step before exposing device to the user space in order to
  avoid races.

In addition clarified description in tpm_chip_register().

Fixes: 313d21ee ("tpm: device class for tpm")
Fixes: afb5abc2 ("tpm: two-phase chip management functions")
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NPeter Huewe <peterhuewe@gmx.de>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

dev_set_name() takes three arguments where the second argument is
a format string. This patch fixes the call accordingly in tpm-chip.c
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Fixes: 313d21ee ("tpm: device class for tpm")
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

TPM 2.0 devices are separated by adding a field 'flags' to struct
tpm_chip and defining a flag TPM_CHIP_FLAG_TPM2 for tagging them.

This patch adds the following internal functions:

- tpm2_get_random()
- tpm2_get_tpm_pt()
- tpm2_pcr_extend()
- tpm2_pcr_read()
- tpm2_startup()

Additionally, the following exported functions are implemented for
implementing TPM 2.0 device drivers:

- tpm2_do_selftest()
- tpm2_calc_ordinal_durations()
- tpm2_gen_interrupt()

The existing functions that are exported for the use for existing
subsystems have been changed to check the flags field in struct
tpm_chip and use appropriate TPM 2.0 counterpart if
TPM_CHIP_FLAG_TPM2 is est.

The code for tpm2_calc_ordinal_duration() and tpm2_startup() were
originally written by Will Arthur.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: NWill Arthur <will.c.arthur@intel.com>
Reviewed-by: NJasob Gunthorpe <jason.gunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NPeter Huewe <peterhuewe@gmx.de>
Tested-by: NPeter Huewe <peterhuewe@gmx.de>

[phuewe: Fixed copy paste error * 2]
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

Added own device class for TPM. Uses MISC_MAJOR:TPM_MINOR for the
first character device in order to retain backwards compatibility.
Added tpm_dev_release() back attached to the character device.

I've been running this code now for a while on my laptop (Lenovo
T430S) TrouSerS works perfectly without modifications. I don't
believe it breaks anything significantly.

The sysfs attributes that have been placed under the wrong place
and are against sysfs-rules.txt should be probably left to
stagnate under platform device directory and start defining
new sysfs attributes to the char device directory.

Guidelines for future TPM sysfs attributes should be probably
along the lines of

- Single flat set of mandatory sysfs attributes. For example,
  current PPI interface is way way too rich when you only want
  to use it to clear and activate the TPM.

- Define sysfs attribute if and only if there's no way to get
  the value from ring-3. No attributes for TPM properties. It's
  just unnecessary maintenance hurdle that we don't want.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NJasob Gunthorpe <jason.gunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Tested-by: NScot Doyle <lkml14@scotdoyle.com>
Tested-by: NPeter Huewe <peterhuewe@gmx.de>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

Rename chip->dev to chip->pdev to make it explicit that this not the
character device but actually represents the platform device.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NJasob Gunthorpe <jason.gunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Reviewed-by: NPeter Huewe <peterhuewe@gmx.de>
Tested-by: NScot Doyle <lkml14@scotdoyle.com>
Tested-by: NPeter Huewe <peterhuewe@gmx.de>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

Traversal of the ACPI device tree was not done right. PPI interface
should be looked up only from the ACPI device that is the platform
device for the TPM. This could cause problems with systems with
two TPM chips such as 4th gen Intel systems.

In addition, added the missing license and copyright platter to
the tpm_ppi.c.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NJasob Gunthorpe <jason.gunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Tested-by: NScot Doyle <lkml14@scotdoyle.com>
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>

tpm_register_hardware() and tpm_remove_hardware() are called often
before initializing the device. The problem is that the device might
not be fully initialized when it comes visible to the user space.

This patch resolves the issue by diving initialization into two
parts:

- tpmm_chip_alloc() creates struct tpm_chip.

- tpm_chip_register() sets up the character device and sysfs
  attributes.

The framework takes care of freeing struct tpm_chip by using the devres
API. The broken release callback has been wiped. ACPI drivers do not
ever get this callback.

Regards to Jason Gunthorpe for carefully reviewing this part of the
code.
Signed-off-by: NJarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: NJasob Gunthorpe <jason.gunthorpe@obsidianresearch.com>
Reviewed-by: NStefan Berger <stefanb@linux.vnet.ibm.com>
Tested-by: NScot Doyle <lkml14@scotdoyle.com>
Tested-by: NPeter Huewe <peterhuewe@gmx.de>
[phuewe: update to upstream changes]
Signed-off-by: NPeter Huewe <peterhuewe@gmx.de>