There's very little need for most of the callsites to get
tp->xmit_goal_size updated. That will cost us divide as is,
so slice the function in two. Also, the only users of the
tp->xmit_goal_size are directly behind tcp_current_mss(),
so there's no need to store that variable into tcp_sock
at all! The drop of xmit_goal_size currently leaves 16-bit
hole and some reorganization would again be necessary to
change that (but I'm aiming to fill that hole with u16
xmit_goal_size_segs to cache the results of the remaining
divide to get that tso on regression).
Bring xmit_goal_size parts into tcp.c
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Evgeniy Polyakov <zbr@ioremap.net>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems that no variables clash such that we couldn't do
the check just once later on. Therefore move it.
Also kill dead obvious comment, dead argument and add
unlikely since this mtu probe does not happen too often.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Wow, it was quite tricky to merge that stream of negations
but I think I finally got it right:
check & replace_ts_recent:
(s32)(rcv_tsval - ts_recent) >= 0 => 0
(s32)(ts_recent - rcv_tsval) <= 0 => 0
discard:
(s32)(ts_recent - rcv_tsval) > TCP_PAWS_WINDOW => 1
(s32)(ts_recent - rcv_tsval) <= TCP_PAWS_WINDOW => 0
I toggled the return values of tcp_paws_check around since
the old encoding added yet-another negation making tracking
of truth-values really complicated.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
I've already forgotten what for this was necessary, anyway
it's no longer used (if it ever was).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the pure assignment case, the earlier zeroing is
still in effect.
David S. Miller raised concerns if the ifs are there to avoid
dirtying cachelines. I came to these conclusions:
> We'll be dirty it anyway (now that I check), the first "real" statement
> in tcp_rcv_established is:
>
> tp->rx_opt.saw_tstamp = 0;
>
> ...that'll land on the same dword. :-/
>
> I suppose the blocks are there just because they had more complexity
> inside when they had to calculate the eff_sacks too (maybe it would
> have been better to just remove them in that drop-patch so you would
> have had less head-ache :-)).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
While looking for a possible reason of bugzilla report on HTB oops:
http://bugzilla.kernel.org/show_bug.cgi?id=12858
I found the code in htb_delete calling htb_destroy_class on zero
refcount is very misleading: it can suggest this is a common path, and
destroy is called under sch_tree_lock. Actually, this can never happen
like this because before deletion cops->get() is done, and after
delete a class is still used by tclass_notify. The class destroy is
always called from cops->put(), so without sch_tree_lock.
This doesn't mean much now (since 2.6.27) because all vulnerable calls
were moved from htb_destroy_class to htb_delete, but there was a bug
in older kernels. The same change is done for other classful scheds,
which, it seems, didn't have similar locking problems here.
Reported-by: m0sia <m0sia@m0sia.ru>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On x86_64, its rather unfortunate that "wait_queue_head_t wait"
field of "struct socket" spans two cache lines (assuming a 64
bytes cache line in current cpus)
offsetof(struct socket, wait)=0x30
sizeof(wait_queue_head_t)=0x18
This might explain why Kenny Chang noticed that his multicast workload
was performing bad with 64 bit kernels, since more cache lines ping pongs
were involved.
This litle patch moves "wait" field next "fasync_list" so that both
fields share a single cache line, to speedup sock_def_readable()
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To mark all features and bugfixes submitted since 4.0.11.
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch enables the load balancing capability of firmware
and hardware to spray traffic into different cpus through
separate rx msix interrupts.
The feature is being enabled for NX3031, NX2031 (old) will be
enabled later. This depends on msi-x and compatibility with
msi and legacy is maintained by enabling single rx ring.
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
o remove max_ prefix from ring sizes, since they don't really
represent max possible sizes.
o cleanup naming of rx ring types (normal, jumbo, lro).
o simplify logic to choose rx ring size, gig ports get half
rx ring of 10 gig ports.
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Detach network interface on PCI suspend and recreate hardware
context after resumes.
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are possible times that a driver may fail to completely initialize,
due to a buggy platform or a buggy kernel. In those cases, we'd rather
fail gracefully instead of a panic. Add a few safety checks to some
critical paths to try and prevent a panic in these corner-case situations.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This cleans up the following pieces of the Rx initialization path:
- Enable the ECC memory fault interrupt in OTHER causes.
- Fix an 82598 initialization of RDRXCTL when depending on RSS and VMDq to
be enabled. We don't need these features enabled to safely set the MVMEN
bit to allow multiple SRRCTL register mappings into the RXDCTL registers.
- Fix the RSS initialization path to not stomp on DCB accidentally. When
configuring the MRQC (multiple Rx queue contol) register, we want to make
sure we only OR in features as necessary, instead of full assignment.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Tx accounting when cleaning during NAPI was not completely properly.
We should use the work_limit to determine when to finish cleaning, and
use the same to return the cleaned status. The impact of running like this
causes the NAPI clean for this Tx to get stuck in a scheduling loop, and
can result in Tx not getting cleaned, ending with a Tx hang and device
reset.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Occasionally if the driver was loaded in a system that
didn't support MSI-X or MSI and was on a shared interrupt,
the driver would then panic in NAPI on the first shared
interrupt because we hadn't called napi_add yet.
Solution: call napi_add before calling request_irq
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The interrupt models using EITR have changed in 82599. The way the register
is laid out, the change is transparent to some of the existing code.
However, some of it isn't. This patch fixes all the cases where EITR
handling is different than 82598.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The rx_no_dma_resources counter reported by ethtool -S ethX is not
counting correctly. In 82599, the queue mappings for the counters need
to be mapped properly, and accounted for properly.
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Acked-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A purely cosmetic change. Report which physical layer is present, instead
of PHY unknown. 82599 added new PHY types for the SFP+ devices, and this
was missed getting updated.
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen
Jarek Poplawski
Eric Dumazet
Dhananjay Phadke
PJ Waskiewicz
Jesse Brandeburg
Alexander Duyck