Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

This commit is similar to 481b8821, and it takes care of removing
references to last_rx. Although this was done in b45d71b7

, we refine
it to take care of RHEL kernel versioning and move the logic to the
beginning of the file.

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

Gwenhael Goavec-Merou authored 6 years ago and Juan David González Cobas committed 6 years ago

since 4.11 pci_enable_msi_exact is no more available and fail with :

spec-sw/kernel/spec-pci.c: In function 'spec_probe':
spec-sw/kernel/spec-pci.c:225:9: error: implicit declaration of function 'pci_enable_msi_exact'; did you mean 'pci_enable_msix_exact'? [-Werror=implicit-function-declaration]
   ret = pci_enable_msi_exact(pdev, 1);
         ^~~~~~~~~~~~~~~~~~~~
         pci_enable_msix_exact
cc1: some warnings being treated as errors
make[5]: *** [/usr/src/linux-headers-4.18.0-2-common/scripts/Makefile.build:323: spec-sw/kernel/spec-pci.o] Error 1
make[4]: *** [/usr/src/linux-headers-4.18.0-2-common/Makefile:1531: _module_spec-sw/kernel] Error 2
make[3]: *** [Makefile:146: sub-make] Error 2
make[2]: *** [Makefile:8: all] Error 2
make[2]: Leaving directory '/usr/src/linux-headers-4.18.0-2-amd64'
make[1]: *** [Makefile:20: all] Error 2
make[1]: Leaving directory 'spec-sw/kernel'
make: *** [Makefile:28: kernel] Error 2

This patch add preprocessor case to detect kernel version.
For kernel > 4.11 pci_alloc_irq_vectors is use instead of pci_enable_msi_exact
and pci_free_irq_vectors instead of pci_disable_msi.

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

Gwenhael Goavec-Merou authored 6 years ago and Juan David González Cobas committed 6 years ago

struct net_device has no more last_rx. According to [1] this line must be
simply supressed for kernel > 4.11 to avoid build failure like :

spec-sw/kernel/wr_nic/nic-core.c: In function '__wrn_rx_descriptor':
spec-sw/kernel/wr_nic/nic-core.c:428:5: error: 'struct net_device' has no member named 'last_rx'
  dev->last_rx = jiffies;
     ^~
make[5]: *** [/usr/src/linux-headers-4.18.0-2-common/scripts/Makefile.build:323: spec-sw/kernel/wr_nic/nic-core.o] Error 1
make[4]: *** [/usr/src/linux-headers-4.18.0-2-common/Makefile:1531: _module_spec-sw/kernel] Error 2
make[3]: *** [Makefile:146: sub-make] Error 2
make[2]: *** [Makefile:8: all] Error 2
make[2]: Leaving directory '/usr/src/linux-headers-4.18.0-2-amd64'
make[1]: *** [Makefile:20: all] Error 2
make[1]: Leaving directory 'spec-sw/kernel'
make: *** [Makefile:28: kernel] Error 2

[1] https://patchwork.ozlabs.org/patch/716831/



Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

Gwenhael Goavec-Merou authored 6 years ago and Juan David González Cobas committed 6 years ago

Since 4.11 signal_pending is declared in linux/sched/signal.h. Using
linux/sched.h result in a build failure :

spec-sw/kernel/wr-nic-dio.c: In function 'wrn_dio_cmd_stamp':
spec-sw/kernel/wr-nic-dio.c:267:7: error: implicit declaration of function 'signal_pending'; did you mean 'timer_pending'? [-Werror=implicit-function-declaration]
   if (signal_pending(current))
       ^~~~~~~~~~~~~~
       timer_pending
cc1: some warnings being treated as errors
make[5]: *** [/usr/src/linux-headers-4.18.0-2-common/scripts/Makefile.build:323: spec-sw/kernel/wr-nic-dio.o] Error 1
make[4]: *** [/usr/src/linux-headers-4.18.0-2-common/Makefile:1531: _module_spec-sw/kernel] Error 2
make[3]: *** [Makefile:146: sub-make] Error 2
make[2]: *** [Makefile:8: all] Error 2
make[2]: Leaving directory '/usr/src/linux-headers-4.18.0-2-amd64'
make[1]: *** [Makefile:20: all] Error 2
make[1]: Leaving directory 'spec-sw/kernel'
make: *** [Makefile:28: kernel] Error 2

This patch adds preprocessor to include correct head according to the kernel version.

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

Gwenhael Goavec-Merou authored 6 years ago and Juan David González Cobas committed 6 years ago

Since kernel 4.15, setup_timer is no more available and compilation fails with :
make -C kernel
make[1]: Entering directory '/.../wr-starting-kit/spec-sw/kernel'
make -C /lib/modules/4.18.0-1-amd64/build M=/.../wr-starting-kit/spec-sw/kernel FMC_BUS_ABS=/.../wr-starting-kit/spec-sw/fmc-bus modules
make[2]: Entering directory '/usr/src/linux-headers-4.18.0-1-amd64'
  CC [M]  /.../wr-starting-kit/spec-sw/kernel/wr_nic/endpoint.o
/.../wr-starting-kit/spec-sw/kernel/wr_nic/endpoint.c: In function 'wrn_ep_open':
/.../wr-starting-kit/spec-sw/kernel/wr_nic/endpoint.c:193:2: error: implicit declaration of function 'setup_timer'; did you mean 'del_timer'? [-Werror=implicit-function-declaration]
  setup_timer(&ep->ep_link_timer, wrn_ep_check_link, timerarg);
  ^~~~~~~~~~~
  del_timer
cc1: some warnings being treated as errors
make[5]: *** [/usr/src/linux-headers-4.18.0-1-common/scripts/Makefile.build:323: /.../wr-starting-kit/spec-sw/kernel/wr_nic/endpoint.o] Error 1
make[4]: *** [/usr/src/linux-headers-4.18.0-1-common/Makefile:1518: _module_/.../wr-starting-kit/spec-sw/kernel] Error 2
make[3]: *** [Makefile:146: sub-make] Error 2
make[2]: *** [Makefile:8: all] Error 2
make[2]: Leaving directory '/usr/src/linux-headers-4.18.0-1-amd64'
make[1]: *** [Makefile:20: all] Error 2
make[1]: Leaving directory '/.../wr-starting-kit/spec-sw/kernel'
make: *** [Makefile:28: kernel] Error 2

This patch add preprocessor case to detect kernel version.
For kernel > 4.15 timer_setup is use instead of setup_timer and callback is
adapted to the new need.

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

The error conditions are very rare and when they happen it means
that we have a bug. So, I do not care much about the possibility
of having many prints in an IRQ handler. Something is really wrong
and we must fix it before releasing.

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

It turns out that trans_start (see 7499f529) was deprecated in RHEL
kernel versions earlier than in mainstream. This adds logic to take care
of all cases.

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

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

The trans_start field from net_device has been removed.

----------------
commit 9b36627acecd5792e81daf1a3bff8eab39ed45fb
Author: Florian Westphal <fw@strlen.de>
Date:   Tue May 3 16:33:14 2016 +0200

    net: remove dev->trans_start
----------------
commit 860e9538a9482bb84589f7d0718a7e6d0a944d58
Author: Florian Westphal <fw@strlen.de>
Date:   Tue May 3 16:33:13 2016 +0200

    treewide: replace dev->trans_start update with helper
----------------

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

The following patch deprecated the macro DEFINE_PCI_DEVICE_TABLE
in favor of using directly the structure type

------------------
commit 92e112fdbb3cb55b43390426501a7efacd893b96
Author: Joe Perches <joe@perches.com>
Date:   Fri Dec 13 11:36:22 2013 -0700

    PCI/checkpatch: Deprecate DEFINE_PCI_DEVICE_TABLE
------------------

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

It should not have any effect on previous kernels

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

Compiling on kernel >=4.5 produces the following error:

  CC [M]  spec-sw/kernel/wr-nic-gpio.o
spec-sw/kernel/wr-nic-gpio.c: In function ‘gc_to_fmc’:
spec-sw/kernel/wr-nic-gpio.c:20:25: error: ‘struct gpio_chip’ has no member named ‘dev’
  struct device *dev = gc->dev;
                         ^
spec-sw/kernel/wr-nic-gpio.c: In function ‘wrn_gpio_init’:
spec-sw/kernel/wr-nic-gpio.c:75:4: error: ‘struct gpio_chip’ has no member named ‘dev’
  gc->dev = &fmc->dev;
    ^

This is due to a change in struct gpio_chip introduced in 4.5,
where the "dev" field was renamed to "parent".

58383c7 gpio: change member .dev to .parent

This patch makes our code compatible with modern kernel versions

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Signed-off-by: Dimitris Lampridis <Dimitris.Lampridis@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

This reverts commit 0598de40.

This was not a memory leakage because the fmc-bus release that
pointer for us. Is it good?

We allocate and it is someone else us free, it lacks of simmetry.
Probably the register function should be called devm_ like all
the others in the kernel who relase for your the memory. In this
way it is clear the behaviour.

We ack the GN4124 **before** the VIC. The other way around makes the
interrupt handling timing dependent. For instance. When the VIC
VIC_CTL_EMU_LEN_W is too short it may happens that the VIC asserts
the interrupt **before** the GN4124 is acked; then, so do not receive
interrupts anymore from the GN4124 because it misses the VIC edge.

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Signed-off-by: Michel Arruat <michel.arruat@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

with -V switch

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Changes in the buildsystem.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>