6d1d8050b4 "block, partition: add partition_meta_info to hd_struct"
added part_unpack_uuid() which assumes that the passed in buffer has
enough space for sprintfing "%pU" - 37 characters including '\0'.
Unfortunately, b5af921ec0 "init: add support for root devices
specified by partition UUID" supplied 33 bytes buffer to the function
leading to the following panic with stackprotector enabled.
Kernel panic - not syncing: stack-protector: Kernel stack corrupted in: ffffffff81b14c7e
[<ffffffff815e226b>] panic+0xba/0x1c6
[<ffffffff81b14c7e>] ? printk_all_partitions+0x259/0x26xb
[<ffffffff810566bb>] __stack_chk_fail+0x1b/0x20
[<ffffffff81b15c7e>] printk_all_paritions+0x259/0x26xb
[<ffffffff81aedfe0>] mount_block_root+0x1bc/0x27f
[<ffffffff81aee0fa>] mount_root+0x57/0x5b
[<ffffffff81aee23b>] prepare_namespace+0x13d/0x176
[<ffffffff8107eec0>] ? release_tgcred.isra.4+0x330/0x30
[<ffffffff81aedd60>] kernel_init+0x155/0x15a
[<ffffffff81087b97>] ? schedule_tail+0x27/0xb0
[<ffffffff815f4d24>] kernel_thread_helper+0x5/0x10
[<ffffffff81aedc0b>] ? start_kernel+0x3c5/0x3c5
[<ffffffff815f4d20>] ? gs_change+0x13/0x13
Increase the buffer size, remove the dangerous part_unpack_uuid() and
use snprintf() directly from printk_all_partitions().
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Szymon Gruszczynski <sz.gruszczynski@googlemail.com>
Cc: Will Drewry <wad@chromium.org>
Cc: stable@vger.kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
BIOS on some HP laptops don't set the speaker-pins as fixed but expose
as jacks, and this confuses the driver as if these pins are
jack-detectable. As a result, the machine doesn't get sounds from
speakers because the driver prepares the power-map update via jack
unsol events which never come up in reality. The bug was introduced
in some time in 3.2 for enabling the power-mapping feature.
This patch fixes the problem by replacing the check of the persistent
power-map bits with a proper is_jack_detectable() call.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=43240
Cc: <stable@vger.kernel.org> [v3.2+]
Signed-off-by: Takashi Iwai <tiwai@suse.de>
V4 updated to current linux-security#next
Targeted for git://gitorious.org/smack-next/kernel.git
Modern application runtime environments like to use
naming schemes that are structured and generated without
human intervention. Even though the Smack limit of 23
characters for a label name is perfectly rational for
human use there have been complaints that the limit is
a problem in environments where names are composed from
a set or sources, including vendor, author, distribution
channel and application name. Names like
softwarehouse-pgwodehouse-coolappstore-mellowmuskrats
are becoming harder to avoid. This patch introduces long
label support in Smack. Labels are now limited to 255
characters instead of the old 23.
The primary reason for limiting the labels to 23 characters
was so they could be directly contained in CIPSO category sets.
This is still done were possible, but for labels that are too
large a mapping is required. This is perfectly safe for communication
that stays "on the box" and doesn't require much coordination
between boxes beyond what would have been required to keep label
names consistent.
The bulk of this patch is in smackfs, adding and updating
administrative interfaces. Because existing APIs can't be
changed new ones that do much the same things as old ones
have been introduced.
The Smack specific CIPSO data representation has been removed
and replaced with the data format used by netlabel. The CIPSO
header is now computed when a label is imported rather than
on use. This results in improved IP performance. The smack
label is now allocated separately from the containing structure,
allowing for larger strings.
Four new /smack interfaces have been introduced as four
of the old interfaces strictly required labels be specified
in fixed length arrays.
The access interface is supplemented with the check interface:
access "Subject Object rwxat"
access2 "Subject Object rwaxt"
The load interface is supplemented with the rules interface:
load "Subject Object rwxat"
load2 "Subject Object rwaxt"
The load-self interface is supplemented with the self-rules interface:
load-self "Subject Object rwxat"
load-self2 "Subject Object rwaxt"
The cipso interface is supplemented with the wire interface:
cipso "Subject lvl cnt c1 c2 ..."
cipso2 "Subject lvl cnt c1 c2 ..."
The old interfaces are maintained for compatibility.
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
Dave Chinner wrote:
> Yes, because you have no idea what the calling context is except
> for the fact that is from somewhere inside filesystem code and the
> filesystem could be holding locks. Therefore, GFP_NOFS is really the
> only really safe way to allocate memory here.
I see. Thank you.
I'm not sure, but can call trace happen where somewhere inside network
filesystem or stackable filesystem code with locks held invokes operations that
involves GFP_KENREL memory allocation outside that filesystem?
----------
[PATCH] SMACK: Fix incorrect GFP_KERNEL usage.
new_inode_smack() which can be called from smack_inode_alloc_security() needs
to use GFP_NOFS like SELinux's inode_alloc_security() does, for
security_inode_alloc() is called from inode_init_always() and
inode_init_always() is called from xfs_inode_alloc() which is using GFP_NOFS.
smack_inode_init_security() needs to use GFP_NOFS like
selinux_inode_init_security() does, for initxattrs() callback function (e.g.
btrfs_initxattrs()) which is called from security_inode_init_security() is
using GFP_NOFS.
smack_audit_rule_match() needs to use GFP_ATOMIC, for
security_audit_rule_match() can be called from audit_filter_user_rules() and
audit_filter_user_rules() is called from audit_filter_user() with RCU read lock
held.
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Casey Schaufler <cschaufler@cschaufler-intel.(none)>
The transmuting directory feature of Smack requires that
the transmuting attribute be explicitly set in all cases.
It seems the users of this facility would expect that the
transmuting attribute be inherited by subdirectories that
are created in a transmuting directory. This does not seem
to add any additional complexity to the understanding of
how the system works.
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
Add support for the IB-NAS6210 and IB-NAS 6220. Describe as much as
currently possible in the devicetree files, including the NAND partitions.
Use the partition scheme of the original firmware by default.
Create a board-ib62x0.c for everything else.
Signed-off-by: Simon Baatz <gmbnomis@gmail.com>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
for cyclic dma, dont makr the descriptor as complte. Fix the remaining users of
cyclic dma which do so
Signed-off-by: Vinod Koul <vinod.koul@linux.intel.com>
Tested-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Tested-by: Mika Westerberg <mika.westerberg@iki.fi>
Add support for Iomega Iconnect system.
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Tested-By: Adam Baker <linux@baker-net.org.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
dns323 and (at least) iconnect platforms are using hw led blinking, so,
instead of having 2 identicals .gpio_blink_set gpio-led hooks, move
dns323 code into gpio.c
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Tested-By: Adam Baker <linux@baker-net.org.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
gpio registers are for 32 gpios. Given that orion_gpio_set_blink is called
directly and not through gpiolib, it needs to make sure that the pin value
given to the internal functions are between 0 and 31.
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Tested-By: Adam Baker <linux@baker-net.org.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Use devicetree to define NAND partitions. Use D-link partition scheme by
default, to be vaguely compatible with their userland.
Changes since last submission (V4):-
* Don't add NAND support then throw it away immediately after [Grant Likely]
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: Jamie Lentin <jm@lentin.co.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Add default configuration for NAND, to be enabled in your board config. Ensure
clock gating is set appropriately when the NAND is enabled.
Acked-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: Jamie Lentin <jm@lentin.co.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Allow a NAND chip using the orion_nand driver to be described using devicetree.
Changes since last submission (V4) [Addressing comments by]:-
* WARN when bank-width is out of range [Andrew Lunn]
Changes since last submission (V3):-
* Document all parameters [Grant Likely]
* Convert bank-width to be in bytes
* Add explicit defaults for cle, ale and bank-width
Signed-off-by: Jamie Lentin <jm@lentin.co.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Add support for the DNS-320 and DNS-325. Describe as much as currently possible
in the devicetree files, create a board-dnskw.c for everything else.
Changes since last submission (V3) [Addressing comments by]:-
* One MACH_DLINK_KIRKWOOD_DT for all dtb files [Grant Likely, Jason Cooper]
* Drop brain-dead select "select CONFIG_MTD_OF_PARTS" [Grant Likely]
* Don't add NAND support then throw it away immediately after [Grant Likely]
* Describe purpose of MPP 41, 42 & 49
Changes since last submission (V2):-
* Use IEEE-compliant "okay", rather than "ok" [Scott Wood]
Signed-off-by: Jamie Lentin <jm@lentin.co.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
When checking capabilities, the question we want to be asking is "does
current() have the capability in the child's namespace?"
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
The pathname of /usr/sbin/tomoyo-editpolicy seen from Ubuntu 12.04 Live CD is
squashfs:/usr/sbin/tomoyo-editpolicy rather than /usr/sbin/tomoyo-editpolicy .
Therefore, we need to accept manager programs which do not start with / .
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Morris <james.l.morris@oracle.com>
Upon resume from standby, there is a possible interrupt
unsafe locking scenario raised when configure the Kernel
with CONFIG_PROVE_LOCKING. So this patch fixes that in
PM driver stuff by calling lock/unlock_irqsave/restore.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In mixed burst (MB) mode, the AHB master always initiates
the bursts with fixed-size when the DMA requests transfers
of size less than or equal to 16 beats.
This patch adds the MB support and the flag that can be
passed from the platform to select it.
MB mode can also give some benefits in terms of performances
on some platforms.
v2: fixed Coding Style
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is to extend the number of MAC address registers
for 16 to 32. In fact, other new 16 registers are available in new
chips and this can help on perfect filter mode for unicast.
This patch also fixes the perfect filtering mode by setting the
bit 31 in the MAC address registers.
v2: fixed Coding Style.
Signed-off-by: Gianni Antoniazzi <gianni.antoniazzi-ext@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
wl_device_dealloc() dereferences the "dev" paramter, so let's move it
under the check for NULL.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some coldfire boards (ie m5253demo) have a dm9000 onboard.
Signed-off-by: Steven King <sfking@fdwdc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David pointed out gcc might generate poor code with 31bit fields.
Using u16 is more than enough and permits a better code output.
Also make the code intent more readable using constants, fixed point arithmetic
not being trivial for everybody.
Suggested-by: David Miller <davem@davemloft.net>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* an improvement to avoid to linearise the whole received packet when not needed
* an improvement for client traffic rerouting after roaming
* a fix for the local translation table state-machine
* minor cleanups and fixes
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
iQEcBAABAgAGBQJPsK8NAAoJEFMQTLzJFOZFBCQIAKD9J3BtxOxN5yJkqNu7vEpT
+67KkZiBu5lnVkOT5NX3g/Q9PUCGHUZFE0elOGn5fxLRO2XsUeZuHb1Bm71V9Ogi
WLnxhCYg1c6yHsrMVqY4czNZrE172/r0BMB9gR4ChngYSpzcx2UH7q8MA7eK7F97
hJRAIw7ySX5fPWwdW3VHxqZjc6YhfH8ZzN9FvM8vJYTaP7zZDn8w3himuGhWceI6
nLhdJUIXR5GsH+ImpgFm/+8apEjsvLnPOnRaMfAXyLbwOmzzfO0BAfTMaw6qr/Al
nZ9f1QDS6yEt+iAyMlQOWYICnSk3RsJSfAJxWLdhnBMaMHsAKA3LyQnBLiVOBB8=
=w/0g
-----END PGP SIGNATURE-----
Merge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge
Included changes:
* an improvement to avoid to linearise the whole received packet when not needed
* an improvement for client traffic rerouting after roaming
* a fix for the local translation table state-machine
* minor cleanups and fixes
This patch addresses a bug in a special case for target core SPC-2 RELEASE
logic where the same physical client (eg: iSCSI InitiatorName) with
differing iSCSI session identifiers (ISID) is allowed to incorrectly release
the same client's SPC-2 reservation from the non reservation holding path.
Note this bug is specific to iscsi-target w/ SPC-2 reservations, and
with the default enforce_pr_isids=1 device attr setting in target-core
controls if a InitiatorName + different ISID reservations are handled
the same as a single iSCSI client entity.
Signed-off-by: Bernhard Kohl <bernhard.kohl@gmx.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Section 15.3.1.2 of the software developer manual has this to say about the
RIPV bit in the IA32_MCG_STATUS register:
RIPV (restart IP valid) flag, bit 0 — Indicates (when set) that program
execution can be restarted reliably at the instruction pointed to by the
instruction pointer pushed on the stack when the machine-check exception
is generated. When clear, the program cannot be reliably restarted at
the pushed instruction pointer.
We need to save the state of this bit in do_machine_check() and use it
in mce_notify_process() to force a signal; even if memory_failure() says
it made a complete recovery ... e.g. replaced a clean LRU page.
Acked-by: Borislav Petkov <bp@amd64.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>
worker_enter_idle() has WARN_ON_ONCE() which triggers if nr_running
isn't zero when every worker is idle. This can trigger spuriously
while a cpu is going down due to the way trustee sets %WORKER_ROGUE
and zaps nr_running.
It first sets %WORKER_ROGUE on all workers without updating
nr_running, releases gcwq->lock, schedules, regrabs gcwq->lock and
then zaps nr_running. If the last running worker enters idle
inbetween, it would see stale nr_running which hasn't been zapped yet
and trigger the WARN_ON_ONCE().
Fix it by performing the sanity check iff the trustee is idle.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
Fix the following build error:
net/sched/sch_fq_codel.c: In function 'fq_codel_dump_stats':
net/sched/sch_fq_codel.c:464:3: error: unknown field 'qdisc_stats' specified in initializer
net/sched/sch_fq_codel.c:464:3: warning: missing braces around initializer
net/sched/sch_fq_codel.c:464:3: warning: (near initialization for 'st.<anonymous>')
net/sched/sch_fq_codel.c:465:3: error: unknown field 'qdisc_stats' specified in initializer
net/sched/sch_fq_codel.c:465:3: warning: excess elements in struct initializer
net/sched/sch_fq_codel.c:465:3: warning: (near initialization for 'st')
net/sched/sch_fq_codel.c:466:3: error: unknown field 'qdisc_stats' specified in initializer
net/sched/sch_fq_codel.c:466:3: warning: excess elements in struct initializer
net/sched/sch_fq_codel.c:466:3: warning: (near initialization for 'st')
net/sched/sch_fq_codel.c:467:3: error: unknown field 'qdisc_stats' specified in initializer
net/sched/sch_fq_codel.c:467:3: warning: excess elements in struct initializer
net/sched/sch_fq_codel.c:467:3: warning: (near initialization for 'st')
make[1]: *** [net/sched/sch_fq_codel.o] Error 1
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
m68k allmodconfig:
net/sched/sch_codel.c: In function ‘dequeue’:
net/sched/sch_codel.c:70: error: implicit declaration of function ‘prefetch’
make[1]: *** [net/sched/sch_codel.o] Error 1
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use DT_MACHINE_START() on the r8a7740 based armadillo 800 eva board.
Also include a tiny DTS file to describe the board and update the
Kconfig dependencies to select CONFIG_USE_OF.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Use DT_MACHINE_START() on the sh73a0 based KZM9G board.
Also include a tiny DTS file to describe the board and
update the Kconfig dependencies to select CONFIG_USE_OF.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Remove percpu_xxx serial functions, all of them were replaced by
this_cpu_xxx or __this_cpu_xxx serial functions
Signed-off-by: Alex Shi <alex.shi@intel.com>
Acked-by: Christoph Lameter <cl@gentwo.org>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Since percpu_xxx() serial functions are duplicated with this_cpu_xxx().
Removing percpu_xxx() definition and replacing them by this_cpu_xxx()
in code. There is no function change in this patch, just preparation for
later percpu_xxx serial function removing.
On x86 machine the this_cpu_xxx() serial functions are same as
__this_cpu_xxx() without no unnecessary premmpt enable/disable.
Thanks for Stephen Rothwell, he found and fixed a i386 build error in
the patch.
Also thanks for Andrew Morton, he kept updating the patchset in Linus'
tree.
Signed-off-by: Alex Shi <alex.shi@intel.com>
Acked-by: Christoph Lameter <cl@gentwo.org>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
percpu_xxx funcs are duplicated with this_cpu_xxx funcs, so replace
them for further code clean up.
And in preempt safe scenario, __this_cpu_xxx funcs may has a bit
better performance since __this_cpu_xxx has no redundant
preempt_enable/preempt_disable on some architectures.
Signed-off-by: Alex Shi <alex.shi@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
- remove all uses of btfixup header
- remove the btfixup header
- remove the btfixup code
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use sparc_config to hold the last two function pointers. There was no
point generating dedicated _ops structures only for these.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
I ended up renaming set_cpu_int to send_ipi to
be consistent all way around.
send_ipi was moved to the *_smp.c files so
we could call the relevant method direct,
without any _ops indirection.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the module_init/module_exit routines and the associated
struct comedi_drive and struct pci_driver to the end of the
source. This is more typical of how other drivers are written and
removes the need for the forward declarations.
Convert the driver to use the module_comedi_pci_driver() macro
which makes the code smaller and a bit simpler.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Ian Abbott <abbotti@mev.co.uk>
Cc: Mori Hess <fmhess@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Convert the refactored comedi pci drivers to use the
module_comedi_pci_driver() macro which makes the code
smaller and a bit simpler.
In the process, rename some of the symbols in the driver
to standardize the comedi pci drivers.
Based on the driver {name}:
1) Rename the comedi_driver to {name}_driver.
2) Use a static string {name} for the comedi_driver.driver_name.
3) Rename the pci probe/remove functions to {name)_pci_probe/remove.
4) Rename the pci id_table to {name}_pci_table.
5) Add the static string {name} for the pci_driver.name.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Ian Abbott <abbotti@mev.co.uk>
Cc: Mori Hess <fmhess@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Introduce the module_comedi_pci_driver macro, and the
associated register/unregister functions, which is a
convenience macro for comedi pci driver modules similar
to module_platform_driver. It is intended to be used by
drivers where the init/exit section does nothing but
register/unregister the comedi driver and associated pci
driver. By using this macro it is possible to eliminate
a few lines of boilerplate code per comedi pci driver.
Add a check to make sure that the pci_driver->name is
set. Once all the comedi pci drivers have been fixed this
will be removed.
Also, when registering the pci driver check for failure
and unregister the comedi driver.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Ian Abbott <abbotti@mev.co.uk>
Cc: Mori Hess <fmhess@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Under most circumstances, the bitmap allocator does not allocate the
same full 24-bit QP number immediately after a QP is destroyed.
This works by using the upper bits of a 24-bit QP number, beyond the
number of QPs that are actually available in the low level driver.
For example, say that the HCA is willing to allocate a maximum of 64K
qps. We use the bits 23..16 as a "counter" which is incremented by 1
at each allocation so that even if the same physical QP is
re-allocated, it will not receive the same 24-bit QP number.
However, we have seen the following scenario:
1. Allocate, say, 255 QPs in succession. This will cause a wrap of the "counter".
2. Destroy the first QP allocated, then allocate a new QP. The new QP,
because of the counter wraparound, will get the same FULL QP number as
the QP just destroyed!
This is a problem because packets in transit can be erroneously
delivered to the new QP when they were meant for the old (destroyed)
QP, because the full QP number of the new QP is identical to the
destroyed QP. (The "counter" mechanism is meant to prevent this by
having the full 24-bit QP numbers differ even if the physical QP on
the HCA is the same. As we see above, however, this mechanism does
not always work).
The best fix for this problem is to allocate QPs in round-robin mode,
so that the physical QP numbers are not immediately re-used.
Found-by: Matthew Finlay <matt@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
