Implement the necessary software segmentation on the normal
TX path so that fast-xmit can use segmentation offload if
the hardware (or driver) supports it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
If drivers want to support S/G (really just gather DMA on TX) then
we can now easily support this on the fast-xmit path since it just
needs to write to the ethernet header (and already has a check for
that being possible.)
However, disallow this on the regular TX path (which has to handle
fragmentation, software crypto, etc.) by calling skb_linearize().
Also allow the related HIGHDMA since that's not interesting to the
code in mac80211 at all anyway.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When we go through the complete TX processing, there are a number
of things like fragmentation and software crypto that require the
checksum to be calculated already.
In favour of maintainability, instead of adding the necessary call
to skb_checksum_help() in all the places that need it, just do it
once before the regular TX processing.
Right now this only affects the TI wlcore and QCA ath10k drivers
since they're the only ones using checksum offload. The previous
commits enabled fast-xmit for them in almost all cases.
For wlcore this even fixes a corner case: when a key fails to be
programmed to hardware software encryption gets used, encrypting
frames with a bad checksum.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The driver can clearly enable fast-xmit since it does rate
control in the device and thus must do duration calculation
there as well.
Acked-by: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The driver can clearly enable fast-xmit since it does rate
control in the device and thus must do duration calculation
there as well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
IBSS can be supported very easily since it uses the standard station
authorization state etc. so it just needs to be covered by the header
building switch statement.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When crypto is offloaded then in some cases it's all handled
by the device, and in others only some space for the IV must
be reserved in the frame. Handle both of these cases in the
fast-xmit path, up to a limit of 18 bytes of space for IVs.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
If the driver handles fragmentation then it wouldn't
be done in software so we can still use the fast-xmit
path in that case.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
For hwsim, the duration field in frames is already not valid for
the common case of HT/VHT MCSes, so there's little point in trying
to keep it accurate for the legacy rates. Enable the fast-xmit code
to allow testing that, although given the dependency on hardware
crypto it will only be enabled in open network configurations.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
In order to speed up mac80211's TX path, add the "fast-xmit" cache
that will cache the data frame 802.11 header and other data to be
able to build the frame more quickly. This cache is rebuilt when
external triggers imply changes, but a lot of the checks done per
packet today are simplified away to the check for the cache.
There's also a more detailed description in the code.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Currently an allyesconfig build [gcc-4.9.1] can generate the following:
WARNING: vmlinux.o(.text.unlikely+0x3864): Section mismatch in
reference from the function cpumask_empty.constprop.3() to the
variable .init.data:nmi_ipi_mask
which comes from the cpumask_empty usage in arch/x86/kernel/nmi_selftest.c.
Normally we would not see a symbol entry for cpumask_empty since it is:
static inline bool cpumask_empty(const struct cpumask *srcp)
however in this case, the variant of the symbol gets emitted when GCC does
constant propagation optimization.
Fix things up so that any locally optimized constprop variants don't warn
when accessing variables that live in the __init sections.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Currently the match() function supports a leading * to match any
prefix and a trailing * to match any suffix. However there currently
is not a combination of both that can be used to target matches of
whole families of functions that share a common substring.
Here we expand the *foo and foo* match to also support *foo* with
the goal of targeting compiler generated symbol names that contain
strings like ".constprop." and ".isra."
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Trying to match the SHT_NUL section isn't useful and causes build failures
on parisc and mn10300 since the addition of section strict white-listing
and __ex_table sanitizing.
Signed-off-by: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Fixes: 050e57fd59 ("modpost: add strict white-listing when referencing....")
Fixes: 52dc0595d5 ("modpost: handle relocations mismatch in __ex_table.")
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
As Guenter pointed out, we were never really calculating the extable entry
size because the pointer arithmetic was simply wrong. We want to check
we're handling the second relocation in __ex_table to infer an entry size,
but we were using (void*) pointers instead of Elf_Rel[a]* ones.
This fixes the problem by moving that check in the caller (since we can
deal with different types of relocations) and add is_second_extable_reloc()
to make the whole thing more readable.
Signed-off-by: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Reported-by: Guenter Roeck <linux@roeck-us.net>
CC: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
As Guenter pointed out, we want to assert that extable_entry_size has been
discovered and not the other way around. Moreover, this sanity check is
only valid when we're not dealing with the first relocation in __ex_table,
since we have not discovered the extable entry size at that point.
This was leading to a divide-by-zero on some architectures and make the
build fail.
Signed-off-by: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Reported-by: Guenter Roeck <linux@roeck-us.net>
CC: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
52dc0595d5 introduced OTHER_TEXT_SECTIONS for identifying what
sections could validly have __ex_table entries. Unfortunately, it
wasn't tested with -ffunction-sections, which some architectures
use.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
32-bit and 64-bit ARM use these sections to store executable code, so
they must be whitelisted in modpost's table of valid text sections.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
DW_DMAC_CORE is slected by PCI or Platform driver, so this symbol shouldn't
be user selectable, so remove the prompt
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Recently Andy changed the 64-bit syscall logic so that
pt_regs->ax is initially set to -ENOSYS, and on syscall exit,
it is updated with the actual return value. This simplified
the logic there.
This patch does the same for 32-bit syscall entry points.
The check for %rax being too big is moved to be just before
the call instruction which dispatches execution through the
syscall table.
There is no way to accidentally skip this check now by jumping
to a label after it. This allows us to remove redundant checks
after ptrace et al.
If %rax is too big, we just skip over the (call, write %rax to
pt_regs->ax) instruction pair. pt_regs->ax remains set to -ENOSYS,
and it gets returned to userspace.
Similar to 64-bit code, this eliminates the "ia32_badsys" code path.
Run-tested.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Alexei Starovoitov <ast@plumgrid.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Will Drewry <wad@chromium.org>
Link: http://lkml.kernel.org/r/1429632194-13445-2-git-send-email-dvlasenk@redhat.com
[ Changelog massage. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This change makes the check exact (no more false positives
on "negative" addresses).
Andy explains:
"Canonical addresses either start with 17 zeros or 17 ones.
In the old code, we checked that the top (64-47) = 17 bits were all
zero. We did this by shifting right by 47 bits and making sure that
nothing was left.
In the new code, we're shifting left by (64 - 48) = 16 bits and then
signed shifting right by the same amount, this propagating the 17th
highest bit to all positions to its left. If we get the same value we
started with, then we're good to go."
While it isn't really important to be fully correct here -
almost all addresses we'll ever see will be userspace ones,
but OTOH it looks to be cheap enough: the new code uses
two more ALU ops but preserves %rcx, allowing to not
reload it from pt_regs->cx again.
On disassembly level, the changes are:
cmp %rcx,0x80(%rsp) -> mov 0x80(%rsp),%r11; cmp %rcx,%r11
shr $0x2f,%rcx -> shl $0x10,%rcx; sar $0x10,%rcx; cmp %rcx,%r11
mov 0x58(%rsp),%rcx -> (eliminated)
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Acked-by: Andy Lutomirski <luto@kernel.org>
Cc: Alexei Starovoitov <ast@plumgrid.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Drewry <wad@chromium.org>
Link: http://lkml.kernel.org/r/1429633649-20169-1-git-send-email-dvlasenk@redhat.com
[ Changelog massage. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This keeps all the related PCI IDs together in the driver where
they are used.
Signed-off-by: Sonny Rao <sonnyrao@chromium.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1429644791-25724-1-git-send-email-sonnyrao@chromium.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This uncore is the same as the Haswell desktop part but uses a
different PCI ID.
Signed-off-by: Sonny Rao <sonnyrao@chromium.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1429569247-16697-1-git-send-email-sonnyrao@chromium.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The core_pmu does not define cpu_* callbacks, which handles
allocation of 'struct cpu_hw_events::shared_regs' data,
initialization of debug store and PMU_FL_EXCL_CNTRS counters.
While this probably won't happen on bare metal, virtual CPU can
define x86_pmu.extra_regs together with PMU version 1 and thus
be using core_pmu -> using shared_regs data without it being
allocated. That could could leave to following panic:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<ffffffff8152cd4f>] _spin_lock_irqsave+0x1f/0x40
SNIP
[<ffffffff81024bd9>] __intel_shared_reg_get_constraints+0x69/0x1e0
[<ffffffff81024deb>] intel_get_event_constraints+0x9b/0x180
[<ffffffff8101e815>] x86_schedule_events+0x75/0x1d0
[<ffffffff810586dc>] ? check_preempt_curr+0x7c/0x90
[<ffffffff810649fe>] ? try_to_wake_up+0x24e/0x3e0
[<ffffffff81064ba2>] ? default_wake_function+0x12/0x20
[<ffffffff8109eb16>] ? autoremove_wake_function+0x16/0x40
[<ffffffff810577e9>] ? __wake_up_common+0x59/0x90
[<ffffffff811a9517>] ? __d_lookup+0xa7/0x150
[<ffffffff8119db5f>] ? do_lookup+0x9f/0x230
[<ffffffff811a993a>] ? dput+0x9a/0x150
[<ffffffff8119c8f5>] ? path_to_nameidata+0x25/0x60
[<ffffffff8119e90a>] ? __link_path_walk+0x7da/0x1000
[<ffffffff8101d8f9>] ? x86_pmu_add+0xb9/0x170
[<ffffffff8101d7a7>] x86_pmu_commit_txn+0x67/0xc0
[<ffffffff811b07b0>] ? mntput_no_expire+0x30/0x110
[<ffffffff8119c731>] ? path_put+0x31/0x40
[<ffffffff8107c297>] ? current_fs_time+0x27/0x30
[<ffffffff8117d170>] ? mem_cgroup_get_reclaim_stat_from_page+0x20/0x70
[<ffffffff8111b7aa>] group_sched_in+0x13a/0x170
[<ffffffff81014a29>] ? sched_clock+0x9/0x10
[<ffffffff8111bac8>] ctx_sched_in+0x2e8/0x330
[<ffffffff8111bb7b>] perf_event_sched_in+0x6b/0xb0
[<ffffffff8111bc36>] perf_event_context_sched_in+0x76/0xc0
[<ffffffff8111eb3b>] perf_event_comm+0x1bb/0x2e0
[<ffffffff81195ee9>] set_task_comm+0x69/0x80
[<ffffffff81195fe1>] setup_new_exec+0xe1/0x2e0
[<ffffffff811ea68e>] load_elf_binary+0x3ce/0x1ab0
Adding cpu_(prepare|starting|dying) for core_pmu to have
shared_regs data allocated for core_pmu. AFAICS there's no harm
to initialize debug store and PMU_FL_EXCL_CNTRS either for
core_pmu.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/20150421152623.GC13169@krava.redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull sparc fixes from David Miller:
1) ldc_alloc_exp_dring() can be called from softints, so use
GFP_ATOMIC. From Sowmini Varadhan.
2) Some minor warning/build fixups for the new iommu-common code on
certain archs and with certain debug options enabled. Also from
Sowmini Varadhan.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
sparc: Use GFP_ATOMIC in ldc_alloc_exp_dring() as it can be called in softirq context
sparc64: Use M7 PMC write on all chips T4 and onward.
iommu-common: rename iommu_pool_hash to iommu_hash_common
iommu-common: fix x86_64 compiler warnings
During some code analysis I realized that atomic_add(), atomic_sub()
and friends are not necessarily inlined AND that each function
is defined multiple times:
atomic_inc: 544 duplicates
atomic_dec: 215 duplicates
atomic_dec_and_test: 107 duplicates
atomic64_inc: 38 duplicates
[...]
Each definition is exact equally, e.g.:
ffffffff813171b8 <atomic_add>:
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 01 3e lock add %edi,(%rsi)
5d pop %rbp
c3 retq
In turn each definition has one or more callsites (sure):
ffffffff81317c78: e8 3b f5 ff ff callq ffffffff813171b8 <atomic_add> [...]
ffffffff8131a062: e8 51 d1 ff ff callq ffffffff813171b8 <atomic_add> [...]
ffffffff8131a190: e8 23 d0 ff ff callq ffffffff813171b8 <atomic_add> [...]
The other way around would be to remove the static linkage - but
I prefer an enforced inlining here.
Before:
text data bss dec hex filename
81467393 19874720 20168704 121510817 73e1ba1 vmlinux.orig
After:
text data bss dec hex filename
81461323 19874720 20168704 121504747 73e03eb vmlinux.inlined
Yes, the inlining here makes the kernel even smaller! ;)
Linus further observed:
"I have this memory of having seen that before - the size
heuristics for gcc getting confused by inlining.
[...]
It might be a good idea to mark things that are basically just
wrappers around a single (or a couple of) asm instruction to be
always_inline."
Signed-off-by: Hagen Paul Pfeifer <hagen@jauu.net>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1429565231-4609-1-git-send-email-hagen@jauu.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We don't use irq_enable_sysexit on 64-bit kernels any more.
Remove all the paravirt and Xen machinery to support it on
64-bit kernels.
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Denys Vlasenko <vda.linux@googlemail.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/8a03355698fe5b94194e9e7360f19f91c1b2cf1f.1428100853.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull networking fixes from David Miller:
"Just a few fixes trickling in at this point.
1) If we see an attached socket on an skb in the ipv4 forwarding path,
bail. This can happen due to races with FIB rule addition, and
deletion, and we should just drop such frames. From Sebastian
Pöhn.
2) pppoe receive should only accept packets destined for this hosts's
MAC address. From Joakim Tjernlund.
3) Handle checksum unwrapping properly in ppp receive properly when
it's encapsulated in UDP in some way, fix from Tom Herbert.
4) Fix some bugs in mv88e6xxx DSA driver resulting from the conversion
from register offset constants to mnenomic macros. From Vivien
Didelot.
5) Fix handling of HCA max message size in mlx4 adapters, from Eran
Ben ELisha"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
net/mlx4_core: Fix reading HCA max message size in mlx4_QUERY_DEV_CAP
tcp: add memory barriers to write space paths
altera tse: Error-Bit on tx-avalon-stream always set.
net: dsa: mv88e6xxx: use PORT_DEFAULT_VLAN
net: dsa: mv88e6xxx: fix setup of port control 1
ppp: call skb_checksum_complete_unset in ppp_receive_frame
net: add skb_checksum_complete_unset
pppoe: Lacks DST MAC address check
ip_forward: Drop frames with attached skb->sk
.config files that do not have MFD_SYSCON set.
Omaps now have a dependency to MFD_SYSCON for system control
module generic register area and some clocks with the changes
done in omap-for-v4.1/prcm-dts branch.
This can be pulled on top of omap-for-v4.1/prcm-dts, or into
fixes for v4.1.
We already do have a slight MFD_SYSCON dependency for
REGULATOR_PBIAS for dual voltage MMC cards on the first MMC
bus for many devices, so from that point of view this can
also be merged separately from omap-for-v4.1/prcm-dts.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=4WLp
-----END PGP SIGNATURE-----
Merge tag 'omap-for-v4.1/prcm-dts-mfd-syscon-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap into next/late
Merge "urgent omap boot fix for v4.1 if MFD_SYSCON is not set" from Tony
Lindgren:
Urgent pull request for v4.1 to booting for custom kernel
.config files that do not have MFD_SYSCON set.
Omaps now have a dependency to MFD_SYSCON for system control
module generic register area and some clocks with the changes
done in omap-for-v4.1/prcm-dts branch.
This can be pulled on top of omap-for-v4.1/prcm-dts, or into
fixes for v4.1.
We already do have a slight MFD_SYSCON dependency for
REGULATOR_PBIAS for dual voltage MMC cards on the first MMC
bus for many devices, so from that point of view this can
also be merged separately from omap-for-v4.1/prcm-dts.
* tag 'omap-for-v4.1/prcm-dts-mfd-syscon-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap:
ARM: OMAP2+: Fix booting with configs that don't have MFD_SYSCON
Signed-off-by: Olof Johansson <olof@lixom.net>
Use list_for_each_entry_safe for iterating because handler may be freed
in the loop.
BUG: unable to handle kernel NULL pointer dereference at 000000000000002c
IP: [<ffffffff814d69c8>] acpi_ec_put_query_handler+0x7/0x1a
Call Trace:
acpi_ec_remove_query_handler+0x87/0x97
acpi_smbus_hc_remove+0x2a/0x44 [sbshc]
acpi_device_remove+0x7b/0x9a
__device_release_driver+0x7e/0x110
driver_detach+0xb0/0xc0
bus_remove_driver+0x54/0xe0
driver_unregister+0x2b/0x60
acpi_bus_unregister_driver+0x10/0x12
acpi_smb_hc_driver_exit+0x10/0x12 [sbshc]
SyS_delete_module+0x1b8/0x210
system_call_fastpath+0x12/0x6a
Signed-off-by: Chris Bainbridge <chris.bainbridge@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If we allocate a seed on behalf ot the user in crypto_rng_reset,
we must ensure that it is zeroed afterwards or the RNG may be
compromised.
Reported-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Now that crypto_rng_reset takes a const argument, we no longer
need to cast away the const qualifier.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Now that all rng implementations have switched over to the new
interface, we can remove the old low-level interface.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch ocnverts the ANSI CPRNG implementation to the new
low-level rng interface.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
The file internal.h is only meant to be used by internel API
implementation and not algorithm implementations. In fact it
isn't even needed here so this patch removes it.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
This patch converts the DRBG implementation to the new low-level
rng interface.
This allows us to get rid of struct drbg_gen by using the new RNG
API instead.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Stephan Mueller <smueller@chronox.de>
This patch adds the helpers that allow the registration and removal
of multiple RNG algorithms.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the function crypto_rng_set_entropy. It is only
meant to be used by testmgr when testing RNG implementations by
providing fixed entropy data in order to verify test vectors.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch converts the low-level crypto_rng interface to the
"new" style.
This allows existing implementations to be converted over one-
by-one. Once that is complete we can then remove the old rng
interface.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
There is no reason why crypto_rng_reset should modify the seed
so this patch marks it as const. Since our algorithms don't
export a const seed function yet we have to go through some
contortions for now.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Pull parisc fixes from Helge Deller:
"The patch by Guenter Roeck fixes the build on parisc which got broken
because of commit f24ffde432 ("parisc: expose number of page table
levels on Kconfig level") and the patch from Matthew Wilcox converts
our code to use the generic scatterlist.h header file"
* 'parisc-4.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
parisc: Replace PT_NLEVELS with CONFIG_PGTABLE_LEVELS
parisc: Eliminate sg_virt_addr() and private scatterlist.h
When array is degraded, read data landed on failed drives will result in
reading rest of data in a stripe. So a single sequential read would
result in same data being read twice.
This patch is to avoid chunk aligned read for degraded array. The
downside is to involve stripe cache which means associated CPU overhead
and extra memory copy.
Test Results:
Following test are done on a enterprise storage node with Seagate 6T SAS
drives and Xeon E5-2648L CPU (10 cores, 1.9Ghz), 10 disks MD RAID6 8+2,
chunk size 128 KiB.
I use FIO, using direct-io with various bs size, enough queue depth,
tested sequential and 100% random read against 3 array config:
1) optimal, as baseline;
2) degraded;
3) degraded with this patch.
Kernel version is 4.0-rc3.
Each individual test I only did once so there might be some variations,
but we just focus on big trend.
Sequential Read:
bs=(KiB) optimal(MiB/s) degraded(MiB/s) degraded-with-patch (MiB/s)
1024 1608 656 995
512 1624 710 956
256 1635 728 980
128 1636 771 983
64 1612 1119 1000
32 1580 1420 1004
16 1368 688 986
8 768 647 953
4 411 413 850
Random Read:
bs=(KiB) optimal(IOPS) degraded(IOPS) degraded-with-patch (IOPS)
1024 163 160 156
512 274 273 272
256 426 428 424
128 576 592 591
64 726 724 726
32 849 848 837
16 900 970 971
8 927 940 929
4 948 940 955
Some notes:
* In sequential + optimal, as bs size getting smaller, the FIO thread
become CPU bound.
* In sequential + degraded, there's big increase when bs is 64K and
32K, I don't have explanation.
* In sequential + degraded-with-patch, the MD thread mostly become CPU
bound.
If you want to we can discuss specific data point in those data. But in
general it seems with this patch, we have more predictable and in most
cases significant better sequential read performance when array is
degraded, and almost no noticeable impact on random read.
Performance is a complicated thing, the patch works well for this
particular configuration, but may not be universal. For example I
imagine testing on all SSD array may have very different result. But I
personally think in most cases IO bandwidth is more scarce resource than
CPU.
Signed-off-by: Eric Mei <eric.mei@seagate.com>
Signed-off-by: NeilBrown <neilb@suse.de>
The default setting of 256 stripe_heads is probably
much too small for many configurations. So it is best to make it
auto-configure.
Shrinking the cache under memory pressure is easy. The only
interesting part here is that we put a fairly high cost
('seeks') on shrinking the cache as the cost is greater than
just having to read more data, it reduces parallelism.
Growing the cache on demand needs to be done carefully. If we allow
fast growth, that can upset memory balance as lots of dirty memory can
quickly turn into lots of memory queued in the stripe_cache.
It is important for the raid5 block device to appear congested to
allow write-throttling to work.
So we only add stripes slowly. We set a flag when an allocation
fails because all stripes are in use, allocate at a convenient
time when that flag is set, and don't allow it to be set again
until at least one stripe_head has been released for re-use.
This means that a spurt of requests will only cause one stripe_head
to be allocated, but a steady stream of requests will slowly
increase the cache size - until memory pressure puts it back again.
It could take hours to reach a steady state.
The value written to, and displayed in, stripe_cache_size is
used as a minimum. The cache can grow above this and shrink back
down to it. The actual size is not directly visible, though it can
be deduced to some extent by watching stripe_cache_active.
Signed-off-by: NeilBrown <neilb@suse.de>
Rather than adjusting max_nr_stripes whenever {grow,drop}_one_stripe()
succeeds, do it inside the functions.
Also choose the correct hash to handle next inside the functions.
This removes duplication and will help with future new uses of
{grow,drop}_one_stripe.
This also fixes a minor bug where the "md/raid:%md: allocate XXkB"
message always said "0kB".
Signed-off-by: NeilBrown <neilb@suse.de>
Depending on the available coding we allow optimized rmw logic for write
operations. To support easier testing this patch allows manual control
of the rmw/rcw descision through the interface /sys/block/mdX/md/rmw_level.
The configuration can handle three levels of control.
rmw_level=0: Disable rmw for all RAID types. Hardware assisted P/Q
calculation has no implementation path yet to factor in/out chunks of
a syndrome. Enforcing this level can be benefical for slow CPUs with
hardware syndrome support and fast SSDs.
rmw_level=1: Estimate rmw IOs and rcw IOs. Execute rmw only if we will
save IOs. This equals the "old" unpatched behaviour and will be the
default.
rmw_level=2: Execute rmw even if calculated IOs for rmw and rcw are
equal. We might have higher CPU consumption because of calculating the
parity twice but it can be benefical otherwise. E.g. RAID4 with fast
dedicated parity disk/SSD. The option is implemented just to be
forward-looking and will ONLY work with this patch!
Signed-off-by: Markus Stockhausen <stockhausen@collogia.de>
Signed-off-by: NeilBrown <neilb@suse.de>
Glue it altogehter. The raid6 rmw path should work the same as the
already existing raid5 logic. So emulate the prexor handling/flags
and split functions as needed.
1) Enable xor_syndrome() in the async layer.
2) Split ops_run_prexor() into RAID4/5 and RAID6 logic. Xor the syndrome
at the start of a rmw run as we did it before for the single parity.
3) Take care of rmw run in ops_run_reconstruct6(). Again process only
the changed pages to get syndrome back into sync.
4) Enhance set_syndrome_sources() to fill NULL pages if we are in a rmw
run. The lower layers will calculate start & end pages from that and
call the xor_syndrome() correspondingly.
5) Adapt the several places where we ignored Q handling up to now.
Performance numbers for a single E5630 system with a mix of 10 7200k
desktop/server disks. 300 seconds random write with 8 threads onto a
3,2TB (10*400GB) RAID6 64K chunk without spare (group_thread_cnt=4)
bsize rmw_level=1 rmw_level=0 rmw_level=1 rmw_level=0
skip_copy=1 skip_copy=1 skip_copy=0 skip_copy=0
4K 115 KB/s 141 KB/s 165 KB/s 140 KB/s
8K 225 KB/s 275 KB/s 324 KB/s 274 KB/s
16K 434 KB/s 536 KB/s 640 KB/s 534 KB/s
32K 751 KB/s 1,051 KB/s 1,234 KB/s 1,045 KB/s
64K 1,339 KB/s 1,958 KB/s 2,282 KB/s 1,962 KB/s
128K 2,673 KB/s 3,862 KB/s 4,113 KB/s 3,898 KB/s
256K 7,685 KB/s 7,539 KB/s 7,557 KB/s 7,638 KB/s
512K 19,556 KB/s 19,558 KB/s 19,652 KB/s 19,688 Kb/s
Signed-off-by: Markus Stockhausen <stockhausen@collogia.de>
Signed-off-by: NeilBrown <neilb@suse.de>
The second and (last) optimized XOR syndrome calculation. This version
supports right and left side optimization. All CPUs with architecture
older than Haswell will benefit from it.
It should be noted that SSE2 movntdq kills performance for memory areas
that are read and written simultaneously in chunks smaller than cache
line size. So use movdqa instead for P/Q writes in sse21 and sse22 XOR
functions.
Signed-off-by: Markus Stockhausen <stockhausen@collogia.de>
Signed-off-by: NeilBrown <neilb@suse.de>