Re: [tpmdd-devel] [PATCH v10 0/8] TPM 2.0 support

[Jarkko Sakkinen]

On Sat, 2014-12-13 at 17:53 +0000, Scot Doyle wrote:
> On Fri, 12 Dec 2014, Jarkko Sakkinen wrote:
> 
> > This patch set enables TPM2 protocol and provides drivers for FIFO and
> > CRB interfaces. This patch set does not export any sysfs attributes for
> > TPM 2.0 because existing sysfs attributes have three non-trivial issues:
> > 
> > - They are associated with the platform device instead of character
> >   device.
> > - They are are not trivial key-value pairs but contain text that is
> >   not easily parsed by a computer.
> > - Raciness as described in
> >   http://kroah.com/log/blog/2013/06/26/how-to-create-a-sysfs-file-correctly/
> > 
> > v2:
> > - Improved struct tpm_chip life-cycle by taking advantage of devres
> >   API.
> > - Refined sysfs attributes as simple key-values thereby not repeating
> >   mistakes in TPM1 sysfs attributes.
> > - Documented functions in tpm-chip.c and tpm2-cmd.c.
> > - Documented sysfs attributes.
> > 
> > v3:
> > - Lots of fixes in calling order in device drivers (thanks to Jason
> >   Gunthorpe for pointing these out!).
> > - Attach sysfs attributes to the misc device because it represents
> >   TPM device to the user space.
> > 
> > v4:
> > - Disable sysfs attibutes for TPM 2.0 for until we can sort out the 
> >   best approach for them.
> > - Fixed all the style issues found with checkpatch.pl.
> > 
> > v5:
> > - missing EXPORT_SYMBOL_GPL()
> > - own class for TPM devices used for TPM 2.0 devices and onwards.
> > 
> > v6:
> > - Non-racy initialization for sysfs attributes using struct device's
> >   groups field.
> > - The class 'tpm' is used now for all TPM devices. For the first device
> >   node major MISC_MAJOR and minor TPM_MINOR is used in order to retain
> >   backwards compatability.
> > 
> > v7:
> > - Release device number and free struct tpm_chip memory inside
> >   tpm_dev_release callback.
> > - Moved code from tpm-interface.c and tpm_dev.c to tpm-chip.c.
> > 
> > v8:
> > - Cleaned up unneeded cast from tpm_transmit_cmd().
> > - Cleaned up redundant PPI_VERSION_LEN constant from tpm_ppi.c.
> > - Fixed tpm_tis to use tpm2_calc_ordinal_duration() for TPM2 devices.
> > - tpm_crb: in crb_recv, check that count can hold the TPM header at
> >   minimum.
> > - tpm_crb: add enumerations for bit flags in start and cancel fields
> >   of the control area.
> > - tpm_crb: use ioremap() for command and response buffer because
> >   they might be anywhere.
> > - tpm_crb: use IO access functions for reading ioremapped buffers
> >   because using direct pointers is not portable.
> > - tpm_crb: only apply ACPI start if start method reported by the
> >   TPM2 ACPI table allows it.
> > - In tpm2_pcr_read() just calculate index and bit and get rid of
> >   hacky loop.
> > - Do not add sysfs attributes for TPM 2.0 devices.
> > 
> > v9:
> > - Fixed compilation issues in v8 (sorry for not using the correct
> >   tree).
> > - Just do "return tpm_chip_register();" instead of copying return
> >   value to a variable.
> > - Removed unused tpm2_startup().
> > - In the CRB driver ACPI TPM2 table could contain platform specific
> >   and therefore inequality test does not work. Fixed in this patch
> >   set.
> > 
> > v10:
> > - Fixed coccicheck and sparse errors and other reported style errors.
> > - Fixed build errors without CONFIG_ACPI.
> > - Fixed build error with CONFIG_OF.
> > - Added TPM_CHIP_FLAG_REGISTERED to mark successful tpm_chip_register().
> >   It is checked in the beginning of tpm_chip_unregister(), which is 
> >   called even when "attach" callback for a device fails because "detach"
> >   callback is always called.
> > - Added TPM_CHIP_FLAG_PPI to mark successful PPI interface lookup because 
> >   in older TPM chips version string might be non-existent.
> > - Check TPM version from the 4th byte of STS register after requesting 
> >   the locality because otherwise the read will return bogus data.
> > - Some TPM chips just give 0xff as the 4th byte so using that for detecting
> >   TPM family is unstable. Instead I chose the approach of using idempotent 
> >   TPM 2.x command to detect such case.
> > 
> > Jarkko Sakkinen (8):
> >   tpm: merge duplicate transmit_cmd() functions
> >   tpm: two-phase chip management functions
> >   tpm: fix raciness of PPI interface lookup
> >   tpm: rename chip->dev to chip->pdev
> >   tpm: device class for tpm
> >   tpm: TPM 2.0 baseline support
> >   tpm: TPM 2.0 CRB Interface
> >   tpm: TPM 2.0 FIFO Interface
> > 
> >  Documentation/ABI/stable/sysfs-class-tpm |  22 +-
> >  drivers/char/tpm/Kconfig                 |   9 +
> >  drivers/char/tpm/Makefile                |   3 +-
> >  drivers/char/tpm/tpm-chip.c              | 256 +++++++++++++
> >  drivers/char/tpm/tpm-dev.c               |  42 +--
> >  drivers/char/tpm/tpm-interface.c         | 263 +++++--------
> >  drivers/char/tpm/tpm-sysfs.c             |  29 +-
> >  drivers/char/tpm/tpm.h                   | 118 +++++-
> >  drivers/char/tpm/tpm2-cmd.c              | 617 +++++++++++++++++++++++++++++++
> >  drivers/char/tpm/tpm_atmel.c             |  25 +-
> >  drivers/char/tpm/tpm_crb.c               | 354 ++++++++++++++++++
> >  drivers/char/tpm/tpm_i2c_atmel.c         |  55 +--
> >  drivers/char/tpm/tpm_i2c_infineon.c      |  43 +--
> >  drivers/char/tpm/tpm_i2c_nuvoton.c       |  71 ++--
> >  drivers/char/tpm/tpm_i2c_stm_st33.c      |  32 +-
> >  drivers/char/tpm/tpm_ibmvtpm.c           |  17 +-
> >  drivers/char/tpm/tpm_infineon.c          |  51 +--
> >  drivers/char/tpm/tpm_nsc.c               |  34 +-
> >  drivers/char/tpm/tpm_ppi.c               | 141 ++++---
> >  drivers/char/tpm/tpm_tis.c               | 224 ++++++-----
> >  drivers/char/tpm/xen-tpmfront.c          |  14 +-
> >  21 files changed, 1820 insertions(+), 600 deletions(-)
> >  create mode 100644 drivers/char/tpm/tpm-chip.c
> >  create mode 100644 drivers/char/tpm/tpm2-cmd.c
> >  create mode 100644 drivers/char/tpm/tpm_crb.c
> > 
> > -- 
> > 2.1.0
> > 
> 
> This patchset works on a TPM 1.2 Toshiba CB35-A3120 using either
> 
> CONFIG_TCG_TPM=y
> CONFIG_TCG_TIS=y
> # CONFIG_TCG_TIS_I2C_ATMEL is not set
> # CONFIG_TCG_TIS_I2C_INFINEON is not set
> # CONFIG_TCG_TIS_I2C_NUVOTON is not set
> # CONFIG_TCG_NSC is not set
> # CONFIG_TCG_ATMEL is not set
> # CONFIG_TCG_INFINEON is not set
> # CONFIG_TCG_CRB is not set
> 
> or
> 
> CONFIG_TCG_TPM=y
> CONFIG_TCG_TIS=y
> CONFIG_TCG_TIS_I2C_ATMEL=y
> CONFIG_TCG_TIS_I2C_INFINEON=y
> CONFIG_TCG_TIS_I2C_NUVOTON=y
> CONFIG_TCG_NSC=y
> CONFIG_TCG_ATMEL=y
> CONFIG_TCG_INFINEON=y
> CONFIG_TCG_CRB=y
> 
> Tested-by: Scot Doyle <lkml14@xxxxxxxxxxxxx>

Thank you so much. I appreciate it. I run it constantly on my T430S.

$ tpm_version 
  TPM 1.2 Version Info:
  Chip Version:        1.2.13.12
  Spec Level:          2
  Errata Revision:     3
  TPM Vendor ID:       STM 
  Vendor Specific data: 50
  TPM Version:         01010000
  Manufacturer Info:   53544d20

For the detection part between 1.2/2.0 we could probably move eventually
to STS3 (with the workaround proposed by Stefan Berger) but I would wait
until TPM2 FIFO modules are available in the market. That's why I chose
bit uglier and more conservative route...

/Jarkko


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/