Mike Frysinger reported that his ptrace testcase showed strange
behaviour on parisc: It was not possible to avoid a syscall and the
return value of a syscall couldn't be changed.
To modify a syscall number, we were missing to save the new syscall
number to gr20 which is then picked up later in assembly again.
The effect that the return value couldn't be changed is a side-effect of
another bug in the assembly code. When a process is ptraced, userspace
expects each syscall to report entrance and exit of a syscall. If a
syscall number was given which doesn't exist, we jumped to the normal
syscall exit code instead of informing userspace that the (non-existant)
syscall exits. This unexpected behaviour confuses userspace and thus the
bug was misinterpreted as if we can't change the return value.
This patch fixes both problems and was tested on 64bit kernel with
32bit userspace.
Signed-off-by: Helge Deller <deller@gmx.de>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: stable@vger.kernel.org # v4.0+
Tested-by: Mike Frysinger <vapier@gentoo.org>
Nikolay Aleksandrov says:
====================
bridge: mcast: add support for temp router port
This set adds support for temporary router port which doesn't depend only
on the incoming queries. It can be refreshed by setting multicast_router to
the same value (3). The first two patches are minor changes that prepare
the code for the third which adds this new type of router port.
In order to be able to dump its information the mdb router port format
is changed in patch 04 and extended similar to how mdb entries format was
done recently.
The related iproute2 changes will be posted if this is accepted.
v2: set val first and adjust router type later in patch 01, patch 03 was
split in 2
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow for more multicast router port information to be dumped such as
timer and type attributes. For that that purpose we need to extend the
MDBA_ROUTER_PORT attribute similar to how it was done for the mdb entries
recently. The new format is thus:
[MDBA_ROUTER_PORT] = { <- nested attribute
u32 ifindex <- router port ifindex for user-space compatibility
[MDBA_ROUTER_PATTR attributes]
}
This way it remains compatible with older users (they'll simply retrieve
the u32 in the beginning) and new users can parse the remaining
attributes. It would also allow to add future extensions to the router
port without breaking compatibility.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for a temporary router port which doesn't depend only on the
incoming query. It can be refreshed if set to the same value, which is
a no-op for the rest.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is needed for the upcoming temporary port router. There's no point
to go through the logic if the value is the same.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using raw values makes it difficult to extend and also understand the
code, give them names and do explicit per-option manipulation in
br_multicast_set_port_router.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Alexandre Belloni says:
====================
phy: micrel: fix issues with interrupt on atmel boards
Since the phy is not polled anymore, there were issues getting a link on the
sama5d* xplained boards.
I'm not too sure about were those fixes should go and I'm wondering whether the
first one shoud be made generic.
For the second one, I found the PHY_HAS_MAGICANEG flag that is not used and I
wondering whether this is related to that kind of issue. I had a quick look at
the history and could'nt find its use.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Disable auto negotiation on init to properly detect an already plugged
cable at boot.
At boot, when the phy is started, it is in the PHY_UP state.
However, if a cable is plugged at boot, because auto negociation is already
enabled at the time we get the first interrupt, the phy is already running.
But the state machine then switches from PHY_UP to PHY_AN and calls
phy_start_aneg(). phy_start_aneg() will not do anything because aneg is
already enabled on the phy. It will then wait for a interrupt before going
further. This interrupt will never happen unless the cable is unplugged and
then replugged.
It was working properly before 321beec504 (net: phy: Use interrupts when
available in NOLINK state) because switching to NOLINK meant starting
polling the phy, even if IRQ were enabled.
Fixes: 321beec504 (net: phy: Use interrupts when available in NOLINK state)
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
At least on ksz8081, when getting back from power down, interrupts are
disabled. ensure they are reenabled if they were previously enabled.
This fixes resuming which is failing on the xplained boards from atmel
since 321beec504 (net: phy: Use interrupts when available in NOLINK
state)
Fixes: 321beec504 (net: phy: Use interrupts when available in NOLINK state)
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David Binderman reported a style issue in the floppy.h header file:
arch/parisc/include/asm/floppy.h:221: (style) Boolean result is used in bitwise
operation. Clarify expression with parentheses.
Reported-by: David Binderman <dcb314@hotmail.com>
Cc: David Binderman <dcb314@hotmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
The two header files got moved to include/linux, and most
users were already converted, this changes the remaining drivers
and removes the files.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Acked-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Yisen Zhuang <yisen.zhuang@huawei.com>
When CONFIG_BUG is disabled, BUG_ON() will only evaluate the condition,
but will not actually stop the current thread. GCC warns about a couple
of BUG_ON() users where this actually leads to further undefined
behavior:
include/linux/ceph/osdmap.h: In function 'ceph_can_shift_osds':
include/linux/ceph/osdmap.h:54:1: warning: control reaches end of non-void function
fs/ext4/inode.c: In function 'ext4_map_blocks':
fs/ext4/inode.c:548:5: warning: 'retval' may be used uninitialized in this function
drivers/mfd/db8500-prcmu.c: In function 'prcmu_config_clkout':
drivers/mfd/db8500-prcmu.c:762:10: warning: 'div_mask' may be used uninitialized in this function
drivers/mfd/db8500-prcmu.c:769:13: warning: 'mask' may be used uninitialized in this function
drivers/mfd/db8500-prcmu.c:757:7: warning: 'bits' may be used uninitialized in this function
drivers/tty/serial/8250/8250_core.c: In function 'univ8250_release_irq':
drivers/tty/serial/8250/8250_core.c:252:18: warning: 'i' may be used uninitialized in this function
drivers/tty/serial/8250/8250_core.c:235:19: note: 'i' was declared here
There is an obvious conflict of interest here: on the one hand, someone
who disables CONFIG_BUG() will want the kernel to be as small as possible
and doesn't care about printing error messages to a console that nobody
looks at. On the other hand, running into a BUG_ON() condition means that
something has gone wrong, and we probably want to also stop doing things
that might cause data corruption.
This patch picks the second choice, and changes the NOP to BUG(), which
normally stops the execution of the current thread in some form (endless
loop or a trap). This follows the logic we applied in a4b5d580e0 ("bug:
Make BUG() always stop the machine").
For ARM multi_v7_defconfig, the size slightly increases:
section CONFIG_BUG=y CONFIG_BUG=n CONFIG_BUG=n+patch
.text 8320248 | 8180944 | 8207688
.rodata 3633720 | 3567144 | 3570648
__bug_table 32508 | --- | ---
__modver 692 | 1584 | 2176
.init.text 558132 | 548300 | 550088
.exit.text 12380 | 12256 | 12380
.data 1016672 | 1016064 | 1016128
Total 14622556 | 14374510 | 14407326
So instead of saving 1.70% of the total image size, we only save 1.48%
by turning off CONFIG_BUG, but in return we can ensure that we don't run
into cases of uninitialized variable or return code uses when something
bad happens. Aside from that, we significantly reduce the number of
warnings in randconfig builds, which makes it easier to fix the warnings
about other problems.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Vivien Didelot says:
====================
net: dsa: mv88e6xxx: implement VLAN filtering
This patchset fixes hardware bridging for non 802.1Q aware systems.
The mv88e6xxx DSA driver currently depends on CONFIG_VLAN_8021Q and
CONFIG_BRIDGE_VLAN_FILTERING enabled for correct bridging between switch ports.
Patch 1/9 adds support for the VLAN filtering switchdev attribute in DSA.
Patchs 2/9 and 3/9 add helper functions for the following patches.
Patchs 4/9 to 6/9 assign dynamic address databases to VLANs, ports, and
bridge groups (the lowest available FID is cleared and assigned), and thus
restore support for per-port FDB operations.
Patchs 7/9 to 9/9 refine ports isolation and setup 802.1Q on user demand.
With this patchset, ports get correctly bridged and the driver behaves as
expected, with or without 802.1Q support.
With CONFIG_VLAN_8021Q enabled, setting a default PVID to the bridge correctly
propagates the corresponding VLAN, in addition to the hardware bridging:
# echo 42 > /sys/class/net/<bridge>/bridge/default_pvid
But considering CONFIG_BRIDGE_VLAN_FILTERING enabled, the hardware VLAN
filtering is enabled on all bridge members only when the user requests it:
# echo 1 > /sys/class/net/<bridge>/bridge/vlan_filtering
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement port_vlan_filtering in the driver to toggle the related port
802.1Q mode between DISABLED and SECURE, on user request.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that ports isolation is correctly configured when joining or leaving
a bridge, there is no need to rely on reserved VLANs to isolate
unbridged ports anymore. Thus remove them, and disable 802.1Q on setup.
This restores the expected behavior of hardware bridging for systems
without 802.1Q or VLAN filtering enabled.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The In Chip Port Based VLAN Table contains bits used to restrict which
output ports this input port can send frames to.
With the VLAN filtering enabled, these tables work in conjunction with
the VLAN Table Unit to allow egressing frames.
In order to remove the current dependency to BRIDGE_VLAN_FILTERING for
basic hardware bridging to work, it is necessary to restore a fine
control of each port's VLANTable, on setup and when a port joins or
leaves a bridge.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Give a new bridge a fresh FDB, assign it to its members, and restore a
fresh FDB to a port leaving a bridge.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Restore per-port FDB. Assign them on setup, allow adding and deleting
addresses into them, and dump them.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a _mv88e6xxx_fid_new function which gives and flushes the lowest FID
available. Call it when preparing a new VTU entry.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move out the code which dumps a single FDB to its own function.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename _mv88e6xxx_vlan_init in _mv88e6xxx_vtu_new, eventually called
from a new _mv88e6xxx_vtu_get function, which abstracts the VTU GetNext
VID-1 trick to retrieve a single entry.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a user explicitly requests VLAN filtering with something like:
# echo 1 > /sys/class/net/<bridge>/bridge/vlan_filtering
Switchdev propagates a SWITCHDEV_ATTR_ID_BRIDGE_VLAN_FILTERING port
attribute.
Add support for it in the DSA layer with a new port_vlan_filtering
function to let drivers toggle 802.1Q filtering on user demand.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ppp_read() and ppp_poll() can be called concurrently with ppp_ioctl().
In this case, ppp_ioctl() might call ppp_ccp_closed(), which may update
ppp->flags while ppp_read() or ppp_poll() is reading it.
The update done by ppp_ccp_closed() isn't atomic due to the bit mask
operation ('ppp->flags &= ~(SC_CCP_OPEN | SC_CCP_UP)'), so concurrent
readers might get transient values.
Reading incorrect ppp->flags may disturb the 'ppp->flags & SC_LOOP_TRAFFIC'
test in ppp_read() and ppp_poll(), which in turn can lead to improper
decision on whether the PPP unit file is ready for reading or not.
Since ppp_ccp_closed() is protected by the Rx and Tx locks (with
ppp_lock()), taking the Rx lock is enough for ppp_read() and ppp_poll()
to guarantee that ppp_ccp_closed() won't update ppp->flags
concurrently.
The same reasoning applies to ppp->n_channels. The 'n_channels' field
can also be written to concurrently by ppp_ioctl() (through
ppp_connect_channel() or ppp_disconnect_channel()). These writes aren't
atomic (simple increment/decrement), but are protected by both the Rx
and Tx locks (like in the ppp->flags case). So holding the Rx lock
before reading ppp->n_channels also prevents concurrent writes.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko says:
====================
Introduce devlink interface and first drivers to use it
There a is need for some userspace API that would allow to expose things
that are not directly related to any device class like net_device of
ib_device, but rather chip-wide/switch-ASIC-wide stuff.
Use cases:
1) get/set of port type (Ethernet/InfiniBand)
2) setting up port splitters - split port into multiple ones and squash again,
enables usage of splitter cable
3) setting up shared buffers - shared among multiple ports within
one chip (work in progress)
4) configuration of switch wide properties - resources division etc - This will
allow to pass configuration that is unacceptable to be passed as
a module option.
First patch of this set introduces a new generic Netlink based interface,
called "devlink". It is similar to nl80211 model and it is heavily
influenced by it, including the API definition. The devlink introduction patch
implements use cases 1) and 2). Other 2 are in development atm and will
be addressed by follow-ups.
It is very convenient for drivers to use devlink, as you can see in other
patches in this set.
Counterpart for devlink is userspace tool for now called "dl". Command line
interface and outputs are derived from "ip" tool so it should be easy
for users to get used to it.
It is available here as a standalone tool for now:
https://github.com/jpirko/devlink
After this is merge in kernel, I will include the "dl" or "devlink" tool
into iproute2 toolset.
Port type setting example:
myhost:~$ dl help
Usage: dl [ OPTIONS ] OBJECT { COMMAND | help }
where OBJECT := { dev | port | monitor }
OPTIONS := { -v/--verbose }
myhost:~$ dl dev help
Usage: dl dev show [DEV]
myhost:~$ dl dev show
pci/0000:01:00.0
myhost:~$ dl port help
Usage: dl port show [DEV/PORT_INDEX]
Usage: dl port set DEV/PORT_INDEX [ type { eth | ib | auto} ]
Usage: dl port split DEV/PORT_INDEX count
Usage: dl port unsplit DEV/PORT_INDEX
myhost:~$ dl port show
pci/0000:01:00.0/1: type ib ibdev mlx4_0
pci/0000:01:00.0/2: type ib ibdev mlx4_0
myhost:~$ sudo dl port set pci/0000:01:00.0/1 type eth
myhost:~$ dl port show
pci/0000:01:00.0/1: type eth netdev ens4
pci/0000:01:00.0/2: type ib ibdev mlx4_0
myhost:~$ sudo dl port set ens4 type auto
myhost:~$ dl port show
pci/0000:01:00.0/1: type eth(auto) netdev ens4
pci/0000:01:00.0/2: type ib ibdev mlx4_0
Port splitting example:
myswitch:~$ sudo modprobe mlxsw_pci
myswitch:~$ dl port
pci/0000:03:00.0/1: type eth netdev eth0
pci/0000:03:00.0/3: type eth netdev eth1
pci/0000:03:00.0/5: type eth netdev eth2
...
pci/0000:03:00.0/63: type eth netdev eth31
myswitch:~$ sudo dl port split pci/0000:03:00.0/1 2 (or "sudo dl port split eth0 2")
myswitch:~$ dl port
pci/0000:03:00.0/3: type eth netdev eth1
pci/0000:03:00.0/5: type eth netdev eth2
...
pci/0000:03:00.0/63: type eth netdev eth31
pci/0000:03:00.0/1: type eth netdev eth0 split_group 16
pci/0000:03:00.0/2: type eth netdev eth32 split_group 16
myswitch:~$ sudo dl port unsplit pci/0000:03:00.0/1
myswitch:~$ dl port
pci/0000:03:00.0/3: type eth netdev eth1
pci/0000:03:00.0/5: type eth netdev eth2
pci/0000:03:00.0/63: type eth netdev eth31
pci/0000:03:00.0/1: type eth netdev eth0
v2->v3:
patch 1/9
-removed generated devlink index and name, use bus name and dev name as
a handle for all userspace originated commands. Along with that,
remove sysfs stub. Requested by Hannes Sowa.
patch 2/9
-add dev param to devlink_register (api change)
patch 4/9
-add dev param to devlink_register (api change)
patch 9/9
-set port's speed according to width fix by Ido
v1->v2:
patch 1/9
-removed no longer used "devlink_dev" helper
-fix couple of typos and misspells
patch 4/9:
-removed SET_NETDEV_DEV set to devlink dev
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow a user to split or unsplit a port using the newly introduced
devlink ops.
Once split, the original netdev is destroyed and 2 or 4 others are
created, according to user configuration. The new ports are like any
other port, with the sole difference of supporting a lower maximum
speed. When unsplit, the reverse process takes place.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When splitting and unsplitting we'll destroy usable ports on the fly, so
mark them using a NULL pointer to indicate that their local port number
is free and can be re-used.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The port netdevs are each associated with a different local port number
in the device. These local ports are grouped into groups of 4 (e.g.
(1-4), (5-8)) called clusters. The cluster constitutes the one of two
possible modules they can be mapped to. This mapping is board-specific
and done by the device's firmware during init.
When splitting a port by 4, the device requires us to first unmap all
the ports in the cluster and then map each to a single lane in the module
associated with the port netdev used as the handle for the operation.
This means that two port netdevs will disappear, as only 100Gb/s (4
lanes) ports can be split and we are guaranteed to have two of these
((1, 3), (5, 7) etc.) in a cluster.
When unsplit occurs we need to reinstantiate the two original 100Gb/s
ports and map each to its origianl module. Therefore, during driver init
store the initial local port to module mapping, so it can be used later
during unsplitting.
Note that a by 2 split doesn't require us to store the mapping, as we
only need to reinstantiate one port whose module is known.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When splitting a port we replace it with 2 or 4 other ports. To be able
to do that we need to remove the original port netdev and unmap it from
its module. However, we first mark it as disabled, as active ports
cannot be unmapped.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add middle layer in mlxsw core code to forward port split/unsplit calls
into specific ASIC drivers.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement newly introduced devlink interface. Add devlink port instances
for every port and set the port types accordingly.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So far, there has been an mlx4-specific sysfs file allowing user to
change port type to either Ethernet of InfiniBand. This is very
inconvenient.
Allow to expose the same ability to set port type in a generic way
using devlink interface.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement newly introduced devlink interface. Add devlink port instances
for every port and set the port types accordingly.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
v2->v3:
-add dev param to devlink_register (api change)
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce devlink infrastructure for drivers to register and expose to
userspace via generic Netlink interface.
There are two basic objects defined:
devlink - one instance for every "parent device", for example switch ASIC
devlink port - one instance for every physical port of the device.
This initial portion implements basic get/dump of objects to userspace.
Also, port splitter and port type setting is implemented.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Calling ib_poll_cq() to sort through WCs during a completion is a
common pattern amongst RDMA consumers. Since commit 14d3a3b249
("IB: add a proper completion queue abstraction"), WC sorting can
be handled by the IB core.
By converting to this new API, svcrdma is made a better neighbor to
other RDMA consumers, as it allows the core to schedule the delivery
of completions more fairly amongst all active consumers.
This new API also aims each completion at a function that is
specific to the WR's opcode. Thus the ctxt->wr_op field and the
switch in process_context is replaced by a set of methods that
handle each completion type.
Because each ib_cqe carries a pointer to a completion method, the
core can now post operations on a consumer's QP, and handle the
completions itself.
The server's rdma_stat_sq_poll and rdma_stat_sq_prod metrics are no
longer updated.
As a clean up, the cq_event_handler, the dto_tasklet, and all
associated locking is removed, as they are no longer referenced or
used.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Calling ib_poll_cq() to sort through WCs during a completion is a
common pattern amongst RDMA consumers. Since commit 14d3a3b249
("IB: add a proper completion queue abstraction"), WC sorting can
be handled by the IB core.
By converting to this new API, svcrdma is made a better neighbor to
other RDMA consumers, as it allows the core to schedule the delivery
of completions more fairly amongst all active consumers.
Because each ib_cqe carries a pointer to a completion method, the
core can now post operations on a consumer's QP, and handle the
completions itself.
svcrdma receive completions no longer use the dto_tasklet. Each
polled Receive WC is now handled individually in soft IRQ context.
The server transport's rdma_stat_rq_poll and rdma_stat_rq_prod
metrics are no longer updated.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Clean up: close_out is reached only when ctxt == NULL and XPT_CLOSE
is already set.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Devesh Sharma <devesh.sharma@broadcom.com>
Tested-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
RFC 5666 Section 4.2 states:
> When the peer detects an RPC-over-RDMA header version that it does
> not support (currently this document defines only version 1), it
> replies with an error code of ERR_VERS, and provides the low and
> high inclusive version numbers it does, in fact, support.
And:
> When other decoding errors are detected in the header or chunks,
> either an RPC decode error MAY be returned or the RPC/RDMA error
> code ERR_CHUNK MUST be returned.
The Linux NFS server does throw ERR_VERS when a client sends it
a request whose rdma_version is not "one." But it does not return
ERR_CHUNK when a header decoding error occurs. It just drops the
request.
To improve protocol extensibility, it should reject invalid values
in the rdma_proc field instead of treating them all like RDMA_MSG.
Otherwise clients can't detect when the server doesn't support
new rdma_proc values.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Devesh Sharma <devesh.sharma@broadcom.com>
Tested-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
When constructing an error reply, svc_rdma_xdr_encode_error()
needs to view the client's request message so it can get the
failing request's XID.
svc_rdma_xdr_decode_req() is supposed to return a pointer to the
client's request header. But if it fails to decode the client's
message (and thus an error reply is needed) it does not return the
pointer. The server then sends a bogus XID in the error reply.
Instead, unconditionally generate the pointer to the client's header
in svc_rdma_recvfrom(), and pass that pointer to both functions.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Devesh Sharma <devesh.sharma@broadcom.com>
Tested-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Fix several issues with svc_rdma_send_error():
- Post a receive buffer to replace the one that was consumed by
the incoming request
- Posting a send should use DMA_TO_DEVICE, not DMA_FROM_DEVICE
- No need to put_page _and_ free pages in svc_rdma_put_context
- Make sure the sge is set up completely in case the error
path goes through svc_rdma_unmap_dma()
- Replace the use of ENOSYS, which has a reserved meaning
Related fixes in svc_rdma_recvfrom():
- Don't leak the ctxt associated with the incoming request
- Don't close the connection after sending an error reply
- Let svc_rdma_send_error() figure out the right header error code
As a last clean up, move svc_rdma_send_error() to svc_rdma_sendto.c
with other similar functions. There is some common logic in these
functions that could someday be combined to reduce code duplication.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Devesh Sharma <devesh.sharma@broadcom.com>
Tested-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Error headers are shorter than either RDMA_MSG or RDMA_NOMSG.
Since HDRLEN_MIN is already used in several other places that would
be annoying to change, add RPCRDMA_HDRLEN_ERR for the one or two
spots where the shorter length is needed.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Clean up: Most svc_rdma_post_recv() call sites close the transport
connection when a receive cannot be posted. Wrap that in a common
helper.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Devesh Sharma <devesh.sharma@broadcom.com>
Tested-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The maximum size of a backchannel message on RPC-over-RDMA depends
on the connection's inline threshold. Today that threshold is
typically 1024 bytes, making the maximum message size 996 bytes.
The Linux server's CREATE_SESSION operation checks that the size
of callback Calls can be as large as 1044 bytes, to accommodate
RPCSEC_GSS. Thus CREATE_SESSION fails if a client advertises the
true message size maximum of 996 bytes.
But the server's backchannel currently does not support RPCSEC_GSS.
The actual maximum size it needs is much smaller. It is safe to
reduce the limit to enable NFSv4.1 on RDMA backchannel operation.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The NFS server's XDR encoders adds an XDR pad for content in the
xdr_buf page list at the beginning of the xdr_buf's tail buffer.
On RDMA transports, Write chunks are sent separately and without an
XDR pad.
If a Write chunk is being sent, strip off the pad in the tail buffer
so that inline content following the Write chunk remains XDR-aligned
when it is sent to the client.
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=294
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
When the Linux NFS server writes an odd-length data item into a
Write chunk, it finishes with XDR pad bytes. If the data item is
smaller than the Write chunk, the pad bytes are written at the end
of the data item, but still inside the chunk (ie, in the
application's buffer). Since this is direct data placement, that
exposes the pad bytes.
XDR pad bytes are inserted in order to preserve the XDR alignment
of the next XDR data item in an XDR stream. But Write chunks do not
appear in the payload XDR stream, and only one data item is allowed
in each chunk. Thus XDR padding is not needed in a Write chunk.
With NFSv4, the Linux NFS server places the results of any
operations that follow an NFSv4 READ or READLINK in the xdr_buf's
tail. Those results also should never be sent as a part of a Write
chunk. The current logic in send_write_chunks() appears to assume
that the xdr_buf's tail contains only pad bytes (ie, NFSv3).
The server should write only the contents of the xdr_buf's page list
in a Write chunk. If there's more than an XDR pad in the tail, that
needs to go inline or in the Reply chunk.
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=294
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The client provides the location of Write chunks into which the
server writes bulk payload. The client provides these when the
Upper Layer Protocol wants direct data placement and the Binding
allows it. (For NFS, this is READ and READLINK operations).
The client also provides the location of a Reply chunk into which
the server writes the non-bulk part of an RPC reply. The client
provides this chunk whenever it believes the reply can be larger
than its receive buffers.
The server then uses the presence of these chunks to determine how
it will form its reply message.
svc_rdma_sendto() was looking for Write and Reply chunks multiple
times for every reply message. It would be more efficient to do it
just once.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The server does indeed now support NFSv4.1 on RDMA transports. It
does not support shifting an RDMA-capable TCP transport (such as
iWARP) to RDMA mode.
Reported-by: Shirley Ma <shirley.ma@oracle.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Remember free allocated client when meeting unsupported state protect how.
Fixes: 50c7b948ad ("nfsd: minor consolidation of mach_cred handling code")
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
John Fastabend says:
====================
tc software only
This adds a software only flag to tc but incorporates a bunch of comments
from the original attempt at this.
First instead of having the offload decision logic be embedded in cls_u32
I lifted into cls_pkt.h so it can be used anywhere and named the flag
TCA_CLS_FLAGS_SKIP_HW (Thanks Jiri ;)
In order to do this I put the flag defines in pkt_cls.h as well. However
it was suggested that perhaps these flags could be lifted into the
upper layer of TCA_ as well but I'm afraid this can not be done with
existing tc design as far as I can tell. The problem is the filters are
packed and unpacked in the classifier specific code and pushing the flags
through the high level doesn't seem easily doable. And we already have
this design where classifiers handle generic options such as actions and
policers. So I think adding one more thing here is OK as 'tc', et. al.
already know how to handle this type of thing.
====================
Acked-by: Pravin B Shelar <pshelar@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
