This way, no external dependency on /acc/... directories exists

Signed-off-by: Juan David Gonzalez Cobas <dcobas@cern.ch>

When registering several mezzanines, the second device_register
causes a reprogram that should not do anything (the bitstream
should not change). The release of fmc->sdb information
must happen before, so that fmc_scan_sdb_tree refills
the new SDB information properly.

This is a kludge over a kludge: actually, fmc_scan_sdb_tree
should be called once in the carrier, so that we do not fall into
this.

This commit reverts 95a811a4, as the correct use of FMC_NO_MEZZANINE
prevents from faking an EEPROM and registering the (physically inexistent)
device

We need to track the commit this binary comes from, as a general
policy

golden has the same bitbanged i2c interface as the fine delay

It is useless and should be no more

This kludge is made to cheat the fmc-bus into the belief
that a valid EEPROM with no FRU information was read,
given that FMC_DEVICE_NO_MEZZANINE is not used in it.

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This reverts commit 5febaac4.

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>