Starting with commit a3bc1f11e9 ("gianfar: Revive SKB
recycling") gianfar driver sooner or later stops transmitting any
packets on SMP machines.
start_xmit() prepares new skb for transmitting, generally it does
three things:
1. sets up all BDs (marks them ready to send), except the first one.
2. stores skb into tx_queue->tx_skbuff so that clean_tx_ring()
would cleanup it later.
3. sets up the first BD, i.e. marks it ready.
Here is what clean_tx_ring() does:
1. reads skbs from tx_queue->tx_skbuff
2. checks if the *last* BD is ready. If it's still ready [to send]
then it it isn't transmitted, so clean_tx_ring() returns.
Otherwise it actually cleanups BDs. All is OK.
Now, if there is just one BD, code flow:
- start_xmit(): stores skb into tx_skbuff. Note that the first BD
(which is also the last one) isn't marked as ready, yet.
- clean_tx_ring(): sees that skb is not null, *and* its lstatus
says that it is NOT ready (like if BD was sent), so it cleans
it up (bad!)
- start_xmit(): marks BD as ready [to send], but it's too late.
We can fix this simply by reordering lstatus/tx_skbuff writes.
Reported-by: Martyn Welch <martyn.welch@ge.com>
Bisected-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Tested-by: Martyn Welch <martyn.welch@ge.com>
Cc: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Cc: Stable <stable@vger.kernel.org> [2.6.33]
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch replaces dev->mc_count in all drivers (hopefully I didn't miss
anything). Used spatch and did small tweaks and conding style changes when
it was suitable.
Jirka
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The gfar_select_queue() function was used to set queue mapping
only for forwarding/bridging applications and the condition
for locally generated packets was completely ignored.
The solution is to remove the gfar_select_queue() function and
use skb_record_rx_queue to set queue mapping for
forwarding/bridging applications. This will ensure that in case of
forwarding/bridging applications txq = rxq will be selected and
skb_tx_hash will be used to pick up a txq for locally generated packets.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch updates the per rx/tx queue stats.
To update the per rx queue stats a new structure has been
introduced rx_q_stats.
The per tx queue stats are updated via the netdev_queue
structure itself.
Note that we update only the tx_packtes, tx_bytes, rx_packets,
rx_bytes and rx_dropped stats on a per queue basis.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to enable filer whenever we need to use multiple RX
queues. Also, need to program RIR0 register with the required
distribution we require, if using RX filer hashing support for
packet distribution to multiple queues.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 46ceb60ca8 ("gianfar: Add
Multiple group Support") introduced the following build error
with CONFIG_NET_POLL_CONTROLLER=y:
CC ggianfar.o
ggianfar.c: In function 'gfar_netpoll':
ggianfar.c:2653: error: invalid storage class for function 'gfar_interrupt'
ggianfar.c:2652: warning: ISO C90 forbids mixed declarations and code
ggianfar.c:2681: error: invalid storage class for function 'adjust_link'
ggianfar.c:2764: error: invalid storage class for function 'gfar_set_multi'
ggianfar.c:2855: error: invalid storage class for function 'gfar_clear_exact_match'
ggianfar.c:2877: error: invalid storage class for function 'gfar_set_hash_for_addr'
ggianfar.c:2898: error: invalid storage class for function 'gfar_set_mac_for_addr'
ggianfar.c:2922: error: invalid storage class for function 'gfar_error'
ggianfar.c:3020: warning: ISO C90 forbids mixed declarations and code
ggianfar.c:3032: error: invalid storage class for function 'gfar_init'
ggianfar.c:3037: error: invalid storage class for function 'gfar_exit'
ggianfar.c:3041: error: initializer element is not constant
ggianfar.c:3042: error: initializer element is not constant
ggianfar.c:3042: warning: ISO C90 forbids mixed declarations and code
ggianfar.c:3042: error: expected declaration or statement at end of input
make[1]: *** [ggianfar.o] Error 1
This patch fixes the issue.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before calling gfar_clean_tx_ring() the driver grabs an irqsave
spinlock, and then tries to recycle skbs. But since
skb_recycle_check() returns 0 with IRQs disabled, we'll never
recycle any skbs.
It appears that gfar_clean_tx_ring() and gfar_start_xmit() are
mostly idependent and can work in parallel, except when they
modify num_txbdfree.
So we can drop the lock from most sections and thus fix the skb
recycling.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
gfar_error() can arrive at the middle of gfar_start_xmit() processing,
and so it can trigger transfers of BDs that we don't yet expect to
be transmitted.
Fix this by locking the tx queues in gfar_error().
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit fba4ed030c ("gianfar: Add Multiple
Queue Support") introduced the following build failure:
CC gianfar.o
gianfar.c: In function 'gfar_restore':
gianfar.c:1249: error: request for member 'napi' in something not a structure or union
This patch fixes the issue.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is OK to poll with disabled IRQs, so remove the warning.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit fba4ed030c ("gianfar: Add Multiple
Queue Support") introduced the following warnings:
CHECK gianfar.c
gianfar.c:333:8: warning: incorrect type in assignment (different address spaces)
gianfar.c:333:8: expected unsigned int [usertype] *baddr
gianfar.c:333:8: got unsigned int [noderef] <asn:2>*<noident>
[... 67 lines skipped ...]
gianfar.c:2565:3: warning: incorrect type in argument 1 (different type sizes)
gianfar.c:2565:3: expected unsigned long const *addr
gianfar.c:2565:3: got unsigned int *<noident>
CC gianfar.o
gianfar.c: In function 'gfar_probe':
gianfar.c:985: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:985: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:993: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:993: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c: In function 'gfar_configure_coalescing':
gianfar.c:1680: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:1680: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:1688: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:1688: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c: In function 'gfar_poll':
gianfar.c:2565: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:2565: warning: passing argument 1 of 'find_next_bit' from incompatible pointer type
gianfar.c:2566: warning: passing argument 2 of 'test_bit' from incompatible pointer type
gianfar.c:2585: warning: passing argument 2 of 'set_bit' from incompatible pointer type
Following warnings left unfixed (looks like sparse doesn't like
locks in loops, so __acquires/__releases() doesn't help):
gianfar.c:441:40: warning: context imbalance in 'lock_rx_qs': wrong count at exit
gianfar.c:441:40: context '<noident>': wanted 0, got 1
gianfar.c:449:40: warning: context imbalance in 'lock_tx_qs': wrong count at exit
gianfar.c:449:40: context '<noident>': wanted 0, got 1
gianfar.c:458:3: warning: context imbalance in 'unlock_rx_qs': __context__ statement expected different context
gianfar.c:458:3: context '<noident>': wanted >= 0, got -1
gianfar.c:466:3: warning: context imbalance in 'unlock_tx_qs': __context__ statement expected different context
gianfar.c:466:3: context '<noident>': wanted >= 0, got -1
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch provides basic hash rules programming via the ethtool
interface.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces multiple group support for etsec2.0
devices.
Multiple group support is provided by mapping the set of enabled
queues to different groups and then programming the per group
regsiters imask, ievent, rstat, tstat.
The queues corresponding to a group are indicated by programming
isrg (interrupt steering) registers.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces multiple Tx and Rx queues.
The incoming packets can be classified into different queues
based on filer rules (out of scope of this patch). The number
of queues enabled will be based on a DTS entries fsl,num_tx_queues
and fsl,num_rx_queues.
Although we are enabling multiple queues, the interrupt coalescing
is on per device level (etsec-1.7 doesn't support multiple rxics
and txics).
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces the group structure. The elements of this
structure are the interrupt lines, their corresponding names,
the register memory map.
The elements for this group are factored out from the gfar_private
structure. The introduction of group structure will help in
providing support for newer versions of etsec.
Currently, the support is present only for single group and
single tx/rx queues.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces per tx and per rx queue structures.
Earlier the members of these structures were inside the
gfar_private structure.
Moving forward if we want to support multiple queues, we need
to refactor the gfar_private structure so that introduction of
multiple queues is easier.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some OF platform drivers are missing module device tables, so they won't
load automatically on boot. This patch fixes the issue by adding proper
MODULE_DEVICE_TABLE() macros to the drivers.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Thanks to various cleanups and refactorings this is now straightforward:
convert the gianfar driver to dev_pm_ops, plus add ->restore() callback
that will fully reinitialize MAC internal registers and BDs.
Note that I kept legacy suspend/resume callbacks so that this patch
doesn't depend on PowerPC changes (i.e. dev_pm_ops support for OF
platform drivers).
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After hibernation we want to just reinitialize BDs, no need to allocate
anything. So, factor out BDs initialization code from
gfar_alloc_skb_resourses().
Also, teach gfar_init_bds() to reuse already allocated RX SKBs, i.e.
just call gfar_init_rxbdp() if a SKB was already allocated and mapped.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We want to just reinitialize RX BDs after hibernation, no need to
map the skb->data again. So let's factor gfar_init_rxbdp() out of
gfar_new_rxbdp().
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For hibernation we want to call gfar_init_mac() without need to
free/allocate_skb_resources sequence, so save the DMA address into a
private struct, and move tbase/rbase initialization to gfar_init_mac().
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Two new functions implemented: gfar_alloc_skb_resources() and
gfar_init_mac(). We'll use gfar_init_mac() for restoring after
hibernation.
The patch just moves the code around, there should be no functional
changes.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
startup_gfar() sets the wrap bit for the last rxbd just after
gfar_new_rxbdp() call, which is issued for all rxbds. And
gfar_new_rxbdp() has the following check already:
if (bdp == priv->rx_bd_base + priv->rx_ring_size - 1)
lstatus |= BD_LFLAG(RXBD_WRAP);
So we don't need to set the bit again.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove dma_free_coherent() from stop_gfar() and gfar_start() calls,
place it into free_skb_resources(). That makes SKB resources management
more understandable, plus free_skb_resources() will be used as a cleanup
routine for gfar_alloc_skb_resources() that will be implemented soon.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We're going to split the startup_gfar() into 3 separate functions,
so let's cleanup the code a little bit so that cosmetic changes
won't distract attention from logical ones.
- Remove needless casts (e.g. (struct sk_buff **)kmalloc());
- Turn 'unsigned long vaddr;' into 'void *vaddr', to avoid casting;
- Add new 'struct device *dev' variable as a shorthand for
'&priv->ofdev->dev' that is used all over the place, also rename
'struct net_device *dev' to 'struct net_device *ndev';
- Turn printk(KERN_ERR ...) to pr_err(...), which is shorter;
- Don't return bogus -1 (i.e. -EPERM) when request_irq() fails;
- Turn '&priv->regs->' to just '®s->'.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (1623 commits)
netxen: update copyright
netxen: fix tx timeout recovery
netxen: fix file firmware leak
netxen: improve pci memory access
netxen: change firmware write size
tg3: Fix return ring size breakage
netxen: build fix for INET=n
cdc-phonet: autoconfigure Phonet address
Phonet: back-end for autoconfigured addresses
Phonet: fix netlink address dump error handling
ipv6: Add IFA_F_DADFAILED flag
net: Add DEVTYPE support for Ethernet based devices
mv643xx_eth.c: remove unused txq_set_wrr()
ucc_geth: Fix hangs after switching from full to half duplex
ucc_geth: Rearrange some code to avoid forward declarations
phy/marvell: Make non-aneg speed/duplex forcing work for 88E1111 PHYs
drivers/net/phy: introduce missing kfree
drivers/net/wan: introduce missing kfree
net: force bridge module(s) to be GPL
Subject: [PATCH] appletalk: Fix skb leak when ipddp interface is not loaded
...
Fixed up trivial conflicts:
- arch/x86/include/asm/socket.h
converted to <asm-generic/socket.h> in the x86 tree. The generic
header has the same new #define's, so that works out fine.
- drivers/net/tun.c
fix conflict between 89f56d1e9 ("tun: reuse struct sock fields") that
switched over to using 'tun->socket.sk' instead of the redundantly
available (and thus removed) 'tun->sk', and 2b980dbd ("lsm: Add hooks
to the TUN driver") which added a new 'tun->sk' use.
Noted in 'next' by Stephen Rothwell.
Reported by Michael Guntsche <mike@it-loops.com>
--------------------
Commit
38bddf04bc gianfar: gfar_remove needs to call unregister_netdev()
breaks the build of the gianfar driver because "dev" is undefined in
this function. To quickly test rc9 I changed this to priv->ndev but I do
not know if this is the correct one.
--------------------
Signed-off-by: David S. Miller <davem@davemloft.net>
If vlan has been enabled. ifdown followed by ifup will lost hardware
related state.
Also remove duplicated operation in gfar_vlan_rx_register().
Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
Acked-by: Dai Haruki <dai.haruki@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit fe192a4911 ("Rework gianfar
driver to use of_mdio infrastructure") the fixed-link support is
broken, the driver oopses at init_phy():
Unable to handle kernel paging request for data at address 0x000000e4
Faulting instruction address: 0xc01cf298
Oops: Kernel access of bad area, sig: 11 [#1]
[...]
NIP [c01cf298] init_phy+0x80/0xdc
LR [c01cf250] init_phy+0x38/0xdc
Call Trace:
[cf81fe80] [c01d1cf8] gfar_enet_open+0x6c/0x19c
[cf81fea0] [c024494c] dev_open+0xfc/0x134
[cf81fec0] [c0242edc] dev_change_flags+0x84/0x1ac
[cf81fee0] [c0399ee0] ic_open_devs+0x168/0x2d8
[cf81ff20] [c039b2e8] ip_auto_config+0x90/0x2a4
[cf81ff60] [c0003884] do_one_initcall+0x34/0x1a8
This patch fixes the oops, and removes phy_node checks, and adds a call
to of_phy_connect_fixed_link() if a phy isn't attached..
Also, remove an old fixed-link code that we don't use any longer.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
The use of it was converted to %pM, but the variable
stuck -- remove it now to not cause spurious warnings.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
alloc_etherdev() used to install default implementations of these
operations, but they must now be explicitly installed in struct
net_device_ops.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 0fd56bb5be ("gianfar:
Add support for skb recycling"), gianfar puts skbuffs that are in
the rx ring back onto the recycle list as-is in case there was a
receive error, but this breaks the following invariant: that all
skbuffs on the recycle list have skb->data = skb->head + NET_SKB_PAD.
The RXBUF_ALIGNMENT realignment done in gfar_new_skb() will be done
twice on skbuffs recycled in this way, causing there not to be enough
room in the skb anymore to receive a full packet, eventually leading
to an skb_over_panic from gfar_clean_rx_ring() -> skb_put().
Resetting the skb->data pointer to skb->head + NET_SKB_PAD before
putting the skb back onto the recycle list restores the mentioned
invariant, and should fix this issue.
Reported-by: Michael Guntsche <mike@it-loops.com>
Tested-by: Michael Guntsche <mike@it-loops.com>
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
cacheable_memzero() is completely overkill for the clearing out the FCB
block which is only 8-bytes. The compiler should easily optimize this
with memset. Additionally, cacheable_memzero() only exists on ppc32 and
thus breaks builds of gianfar on ppc64.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch simplifies the driver by making use of more common code.
Tested on Freescale MPC8349emitxgp eval board
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code that was added to increase headroom was wrong.
It doesn't handle the case where gfar_add_fcb() changes the skb.
Better to do check at start of transmit (outside of lock), where
error handling is better anyway.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Gianfar uses a hardware header FCB for offloading. However when used
with bridging or IP forwarding, TX skb might not have enough headroom
for the FCB. Reallocate skb for such cases.
Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 4826857f1b ("gianfar: pass the
proper dev to DMA ops") introduced this build breakage:
CC drivers/net/gianfar.o
drivers/net/gianfar.c: In function 'gfar_suspend':
drivers/net/gianfar.c:552: error: 'struct gfar_private' has no member named 'dev'
drivers/net/gianfar.c: In function 'gfar_resume':
drivers/net/gianfar.c:601: error: 'struct gfar_private' has no member named 'dev'
make[2]: *** [drivers/net/gianfar.o] Error 1
Fix this by converting suspend and resume routines to use
gfar_private->ndev.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to be passing the of_platform device struct into the DMA ops as
its the one that has the archdata setup to know which low-level DMA ops we
should be using (not the net_device one). This isn't an issue until we
expect the archdata to be setup correctly.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Anton Vorontsov
Jiri Pirko
Sandeep Gopalpet
David S. Miller
Linus Torvalds
Toru UCHIYAMA
Yong Zhang
Johannes Berg
Ben Hutchings
Lennert Buytenhek
Kumar Gala
Grant Likely
Markus Brunner
Li Yang
Stephen Hemminger