percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
If we wait for the once-per-second cleanup to free transmit SKBs,
sockets with small transmit buffer sizes might spend most of their
time blocked waiting for the cleanup.
Normally we do a cleanup for each transmitted packet. We add a
watchdog type timer so that we also schedule a timeout for 150uS after
a packet is transmitted. The watchdog is reset for each transmitted
packet, so for high packet rates, it never expires. At these high
rates, the cleanups are done for each packet so the extra watchdog
initiated cleanups are neither needed nor triggered.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
To: netdev@vger.kernel.org
To: gregkh@suse.de
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Patchwork: http://patchwork.linux-mips.org/patch/968/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This version has spelling and comment changes based on feedback from
Eric Dumazet.
Octeon ethernet hardware can handle NETIF_F_SG, so we enable it.
A gather list of up to six fragments will fit in the SKB's CB
structure, so no extra memory is required. If a SKB has more than six
fragments, we must linearize it.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
To: gregkh@suse.de
Patchwork: http://patchwork.linux-mips.org/patch/838/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Convert the driver to be a reasonably well behaved NAPI citizen.
There is one NAPI instance per CPU shared between all input ports. As
receive backlog increases, NAPI is scheduled on additional CPUs.
Receive buffer refill code factored out so it can also be called from
the periodic timer. This is needed to recover from temporary buffer
starvation conditions.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
To: gregkh@suse.de
Patchwork: http://patchwork.linux-mips.org/patch/839/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Stop the queue if too many packets are queued. Restart it from a high
resolution timer.
Rearrange and simplify locking and SKB freeing code
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
To: gregkh@suse.de
Patchwork: http://patchwork.linux-mips.org/patch/843/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
After aligning the blocks returned by kmalloc, we need to save the original
pointer so they can be correctly freed.
There are no guarantees about the alignment of SKB data, so we need to
handle worst case alignment.
Since right shifts over subtraction have no distributive property, we need
to fix the back pointer calculation.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
Patchwork: http://patchwork.linux-mips.org/patch/884/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
removed needless checks in arlan-main.c and slicoss.c
fixed bug in et131x_netdev.c to actually fill addresses in.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The octeon-ethernet driver shares an mdio bus with the octeon-mgmt
driver. Here we convert the octeon-ethernet driver to use the PHY
Abstraction Layer.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Allocate MAC addresses using the same method as the bootloader. This
avoids changing the MAC between bootloader and kernel operation as
well as avoiding duplicates and use of addresses outside of the
assigned range.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
In addition to being magic numbers, the irq number passed to free_irq
is incorrect. We need to use the correct symbolic value instead.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
If we reach the test just below the loop with a `timeout' value of 0,
this does not mean that the timeout caused the loop to end, but rather
the `smi_rd.s.pending', in the last iteration. If timeout caused the
loop to end, then `timeout' is -1, not 0.
Since this can occur only in the last iteration, it is not very likely
to be a problem. By changing the post- to prefix decrement we ensure
that a timeout of 0 does mean it timed out.
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
No need to put ethtool_ops in data, they should be const.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The existing code had the following race:
Thread-1 Thread-2
inc/read in_use
inc/read in_use
inc tx_free_list[qos].len
inc tx_free_list[qos].len
The actual in_use value was incremented twice, but thread-1 is going
to free memory based on its stale value, and will free one too many
times. The result is that memory is freed back to the kernel while
its packet is still in the transmit buffer. If the memory is
overwritten before it is transmitted, the hardware will put a valid
checksum on it and send it out (just like it does with good packets).
If by chance the TCP flags are clobbered but not the addresses or
ports, the result can be a broken TCP stream.
The fix is to track the number of freed packets in a single location
(a Fetch-and-Add Unit register). That way it can never get out of sync
with itself.
We try to free up to MAX_SKB_TO_FREE (currently 10) buffers at a time.
If fewer are available we adjust the free count with the difference.
The action of claiming buffers to free is atomic so two threads cannot
claim the same buffers.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Convert the driver to use net_device_ops as it is now mandatory.
Also compensate for the removal of struct sk_buff's dst field.
The changes are mostly mechanical, the content of ethernet-common.c
was moved to ethernet.c and ethernet-common.{c,h} are removed.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The octeon-ethernet driver supports the sgmii, rgmii, spi, and xaui
ports present on the Cavium OCTEON family of SOCs. These SOCs are
multi-core mips64 processors with existing support over in arch/mips.
The driver files can be categorized into three basic groups:
1) Register definitions, these are named cvmx-*-defs.h
2) Main driver code, these have names that don't start cvmx-.
3) Interface specific functions and other utility code, names starting
with cvmx-
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
