19 Nov, 2015
1 commit
-
Pull s390 fixes from Martin Schwidefsky:
"Assorted bug fixes, the mlock2 system call gets added, and one
improvement. The boot from dasd devices is now possible from a wider
range of devices"* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
s390: remove SALIPL loader
s390: wire up mlock2 system call
s390: remove g5 elf platform support
s390: avoid cache aliasing under z/VM and KVM
s390/sclp: _sclp_wait_int(): retain full PSW mask
s390/zcrypt: Fix initialisation when zcrypt is built-in
s390/zcrypt: Fix kernel crash on systems without AP bus support
s390: add support for ipl devices in subchannel sets > 0
s390/ipl: fix out of bounds access in scpdata_write
s390/pci_dma: improve debugging of errors during dma map
s390/pci_dma: handle dma table failures
s390/pci_dma: unify label of invalid translation table entries
s390/syscalls: remove system call number calculation
s390/cio: simplify css_generate_pgid
s390/diag: add a s390 prefix to the diagnose trace point
s390/head: fix error message on unsupported hardware
11 Nov, 2015
3 commits
-
ap_bus and zcrypt_api assumed module information to always be present
and initialisation to be done in module loading order (symbol
dependencies). These assumptions don't hold if zcrypt is built-in;
THIS_MODULE will be NULL in this case and init call order is linker
order, i.e. Makefile order.Fix initialisation order by ordering the object files in the Makefile
according to their dependencies, like the module loader would do.Fix message type registration by using a dedicated "name" field rather
than piggy-backing on the module ("owner") information. There's no
change to the requirement that module name and msgtype name are
identical. The existing name macros are used.We don't need any special code for dealing with the drivers being
built-in; the generic module support code already does the right
thing.Test results:
1. CONFIG_MODULES=y, CONFIG_ZCRYPT=yKVM: boots, no /sys/bus/ap (expected)
LPAR with CEX5: boots, /sys/bus/ap/devices/card*/type present2. CONFIG_MODULES=y, CONFIG_ZCRYPT=m=:
KVM: boots, loading zcrypt_cex4 (and ap) fails (expected)
LPAR with CEX5: boots, loading =zcrypt_cex4= succeeds,
/sys/bus/ap/devices/card*/type present after explicit module
loading3. CONFIG_MODULES unset, CONFIG_ZCRYPT=y:
KVM: boots, no /sys/bus/ap (expected)
LPAR with CEX5: boots, /sys/bus/ap/devices/card*/type presentNo further testing (user-space functionality) was done.
Fixes: 3b6245fd303f ("s390/zcrypt: Separate msgtype implementation from card modules.")
Signed-off-by: Sascha Silbe
Signed-off-by: Martin Schwidefsky -
On systems without AP bus (e.g. KVM) the kernel crashes during init
calls when zcrypt is built-in:kernel BUG at drivers/base/driver.c:153!
illegal operation: 0001 ilc:1 [#1] SMP
Modules linked in:
CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.2.0+ #221
task: 0000000010a40000 ti: 0000000010a48000 task.ti:0000000010a48000
Krnl PSW : 0704c00180000000 0000000000592bd6(driver_register+0x106/0x140)
R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:3 CC:0 PM:0 EA:3
0000000000000012 0000000000000000 0000000000c45328 0000000000c44e30
00000000009ef63c 000000000067f598 0000000000cf3c58 0000000000000000
000000000000007b 0000000000cb1030 0000000000000002 0000000000000000
0000000000ca8580 0000000010306700 00000000001001d8 0000000010a4bd88
Krnl Code: 0000000000592bc6: f0b00004ebcf srp 4(12,%r0),3023(%r14),0
0000000000592bcc: f0a0000407f4 srp 4(11,%r0),2036,0
#0000000000592bd2: a7f40001 brc 15,592bd4
>0000000000592bd6: e330d0000004 lg %r3,0(%r13)
0000000000592bdc: c0200021edfd larl %r2,9d07d6
0000000000592be2: c0e500126d8f brasl %r14,7e0700
0000000000592be8: e330d0080004 lg %r3,8(%r13)
0000000000592bee: a7f4ffab brc 15,592b44
Call Trace:
([] do_one_initcall+0x90/0x1d0)
[] kernel_init_freeable+0x1e4/0x2a0
[] kernel_init+0x2a/0x120
[] kernel_thread_starter+0x6/0xc
[] kernel_thread_starter+0x0/0xc
Last Breaking-Event-Address:
[] driver_register+0x102/0x140When zcrypt is built as a module, the module loader ensures that the
driver modules cannot be loaded if the AP bus module returns an error
during initialisation. But if zcrypt and the driver are built-in, the
driver is getting initialised even if the AP bus initialisation
failed. The driver invokes ap_driver_register() during initialisation,
which then causes operations on uninitialised data structures to be
performed.Explicitly protect ap_driver_register() by introducing an
"initialised" flag that gets set iff the AP bus initialisation was
successful. When the AP bus initialisation failed,
ap_driver_register() will error out with -ENODEV, causing the driver
initialisation to fail as well.Test results:
1. Inside KVM (no AP bus), zcrypt built-inBoots. /sys/bus/ap not present (expected).
2. Inside KVM (no AP bus), zcrypt as module
Boots. Loading zcrypt_cex4 fails because loading ap_bus fails
(expected).3. On LPAR with CEX5, zcrypt built-in
Boots. /sys/bus/ap/devices/card* present but .../card*/type missing
(i.e. zcrypt_device_register() fails, unrelated issue).4. On LPAR with CEX5, zcrypt as module
Boots. Loading zcrypt_cex4 successful,
/sys/bus/ap/devices/card*/type present. No further testing
(user-space functionality) was done.Signed-off-by: Sascha Silbe
Signed-off-by: Martin Schwidefsky -
Allow to ipl from CCW based devices residing in any subchannel set.
Reviewed-by: Michael Holzheu
Signed-off-by: Sebastian Ott
Signed-off-by: Martin Schwidefsky
09 Nov, 2015
1 commit
-
Simplify the css_generate_pgid code by using stap() independent of
CONFIG_SMP. For !CONFIG_SMP builds stap() will deliver the address
of the cpu we IPL'ed from (which can be != 0).Note: the ifdef was likely added to be compatible with _very_ old
machines which we don't support anyway.Reviewed-by: Peter Oberparleiter
Signed-off-by: Sebastian Ott
Signed-off-by: Martin Schwidefsky
08 Nov, 2015
1 commit
-
No functional changes in this patch, but it prepares us for returning
a more useful cookie related to the IO that was queued up.Signed-off-by: Jens Axboe
Acked-by: Christoph Hellwig
Acked-by: Keith Busch
05 Nov, 2015
1 commit
-
Pull s390 updates from Martin Schwidefsky:
"There is only one new feature in this pull for the 4.4 merge window,
most of it is small enhancements, cleanup and bug fixes:- Add the s390 backend for the software dirty bit tracking. This
adds two new pgtable functions pte_clear_soft_dirty and
pmd_clear_soft_dirty which is why there is a hit to
arch/x86/include/asm/pgtable.h in this pull request.- A series of cleanup patches for the AP bus, this includes the
removal of the support for two outdated crypto cards (PCICC and
PCICA).- The irq handling / signaling on buffer full in the runtime
instrumentation code is dropped.- Some micro optimizations: remove unnecessary memory barriers for a
couple of functions: [smb_]rmb, [smb_]wmb, atomics, bitops, and for
spin_unlock. Use the builtin bswap if available and make
test_and_set_bit_lock more cache friendly.- Statistics and a tracepoint for the diagnose calls to the
hypervisor.- The CPU measurement facility support to sample KVM guests is
improved.- The vector instructions are now always enabled for user space
processes if the hardware has the vector facility. This simplifies
the FPU handling code. The fpu-internal.h header is split into fpu
internals, api and types just like x86.- Cleanup and improvements for the common I/O layer.
- Rework udelay to solve a problem with kprobe. udelay has busy loop
semantics but still uses an idle processor state for the wait"* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (66 commits)
s390: remove runtime instrumentation interrupts
s390/cio: de-duplicate subchannel validation
s390/css: unneeded initialization in for_each_subchannel
s390/Kconfig: use builtin bswap
s390/dasd: fix disconnected device with valid path mask
s390/dasd: fix invalid PAV assignment after suspend/resume
s390/dasd: fix double free in dasd_eckd_read_conf
s390/kernel: fix ptrace peek/poke for floating point registers
s390/cio: move ccw_device_stlck functions
s390/cio: move ccw_device_call_handler
s390/topology: reduce per_cpu() invocations
s390/nmi: reduce size of percpu variable
s390/nmi: fix terminology
s390/nmi: remove casts
s390/nmi: remove pointless error strings
s390: don't store registers on disabled wait anymore
s390: get rid of __set_psw_mask()
s390/fpu: split fpu-internal.h into fpu internals, api, and type headers
s390/dasd: fix list_del corruption after lcu changes
s390/spinlock: remove unneeded serializations at unlock
...
03 Nov, 2015
5 commits
-
cio_validate_io_subchannel() and cio_validate_msg_subchannel() are
identical, as the called functions already take care about the
differences between subchannel types.Just inline the code into the only user,
cio_validate_subchannel(), instead.Signed-off-by: Pierre Morel
Reviewed-by: Cornelia Huck
Acked-by: Sebastian Ott
Signed-off-by: Martin Schwidefsky -
The ret variable is always set by the fn function.
There is no need to initialize it.Signed-off-by: Pierre Morel
Reviewed-By: Sascha Silbe
Acked-by: Cornelia Huck
Signed-off-by: Martin Schwidefsky -
Path verification is either done via dasd_eckd_read_conf() which is
triggered during online processing and resume or via
do_path_verification_work() which is triggered after path events.
The dasd_eckd_read_conf() version added paths unconditionally and did
not check if the path mask was empty. This led to devices having the
disconnected stop flag set but a valid path mask. So they where not
working although they had paths validated successfully. After a resume
this state could even not be solved with additional paths added.Fix by checking for an empty path mask in dasd_eckd_read_conf() and
clearing the device stop bits for a newly added channel path.Reviewed-by: Sebastian Ott
Signed-off-by: Stefan Haberland
Signed-off-by: Martin Schwidefsky -
For a valid PAV assignment the DASD driver needs to notice possibly
changed configuration data. Thus the failing of read configuration
data should also fail the device restore to prevent invalid PAV
assignment. The failed device may get restored after additional paths
get available later on.
If the restore fails after the device was added to the lcu alias
handling it needs to be removed from the alias handling before exiting
the restore function.Reviewed-by: Sebastian Ott
Signed-off-by: Stefan Haberland
Signed-off-by: Martin Schwidefsky -
The configuration data is stored per path and also the first valid
configuration data per device. When dasd_eckd_read_conf is called
again after a path got lost the device configuration data is cleared
but possibly not the per path configuration data. This might lead to a
double free when the lost path gets operational again.Fix by clearing all per path configuration data when the first valid
configuration data is received and stored.Reviewed-by: Sebastian Ott
Signed-off-by: Stefan Haberland
Signed-off-by: Martin Schwidefsky
27 Oct, 2015
3 commits
-
device_ops.c should only contain functions that are called by ccw device
drivers. Move the cio internal functions that handle unconditional
reserve + release to device_pgid.cAcked-by: Peter Oberparleiter
Signed-off-by: Sebastian Ott
Signed-off-by: Martin Schwidefsky -
device_ops.c should only contain functions that are called by ccw device
drivers. Move the cio internal function ccw_device_call_handler to
device_fsm.c where it's used. Remove some useless comments while at it.Acked-by: Peter Oberparleiter
Signed-off-by: Sebastian Ott
Signed-off-by: Martin Schwidefsky -
Remove all the casts to and from the machine check interruption code.
This patch changes struct mci to a union, which contains an anonymous
structure with the already known bits and in addition an unsigned
long field, which contains the raw machine check interruption code.This allows to simply assign and decoce the interruption code value
without the need for all those casts we had all the time.Signed-off-by: Heiko Carstens
Signed-off-by: Martin Schwidefsky
15 Oct, 2015
1 commit
-
A summary unit check occurs when the lcu updates the PAV configuration
e.g. base PAV assignment or PAV mode at all. This requires the reset
of the drivers internal pavgroups. Therefore the alias devices are
flushed and moved via a temporary list to the active_devices list
where they are not associated with a pavgroup. In conjunction with
updates to the base device the pavgroup may be removed since both
base_list and alias_list are empty. Unfortunately during alias flush
and move to the active_device list from alias_list the pavgroup
pointer is not deleted in the device private structure. This leads to
a list del_corruption if another lcu_update tries to move the device
in the non existent pavgroup.Fix by removing the pavgroup pointer after the alias device was moved
to the active_devices list.Signed-off-by: Stefan Haberland
Signed-off-by: Martin Schwidefsky
14 Oct, 2015
21 commits
-
There is a system work queue system_long_wq for long running work.
Use this work queue for the AP bus scan loop.Reviewd-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
Remove the code for really old crypt cards, PCICC and PCICA.
These cards have been out of service for several years.Reviewd-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
Replace the two fields 'unregistered' and 'reset' with a device
state with 5 possible values. Introduce two events for the AP devices,
device poll and device timeout. With the state machine it is easier
to deal with device initialization and suspend/resume. Device polling
is simpler as well, the arkane 'flags' passing is gone.Reviewd-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
If a AP device is removed while messages are still pending, the requests
are cancelled by calling the message receive function with an error pointer
for the reply. The message type receive handler recognize this and create
a fake hardware error TYPE82_RSP_CODE / REP82_ERROR_MACHINE_FAILURE.
The message with the hardware error then causes a printk and a return
code of -EAGAIN.Replace the intricate scheme with an explicit return code for this sitation
and avoid the error message.Reviewd-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
Set the configuration timer at the end of the ap_scan_bus function.
Make use of setup_timer and remove some unnecessary add_timer, mod_timer
and del_timer_sync calls. Replace the complicated timer_pending, mod_timer
and add_timer code in ap_config_time_store with a simple mod_timer.Reviewd-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
If there are no devices on the AP bus there will not be a single
call to the per-device ap_bus_suspend function. Even worse,
there will not be a call to the per-device ap_bus_resume either
and the AP will fail so resume correctly.Introduce a bus specific dev_pm_ops to suspend / resume the AP
bus related things. While we are at it, simplify the power management
code of the AP bus.Reviewd-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
The ap_queue_messsage function will call device_unregister if the
unregistered field of the device has been set while trying to queue
a message. This races with other device_unregister calls, e.g. from
the ap_scan_bus. Remove the call to device_unregister from
ap_queue_message and let ap_scan_bus deal with it.Signed-off-by: Martin Schwidefsky
-
The ap_query_configuration function allocates the ap_config_info
structure, but there is no code to free the structure.
Allocate the structure in the module_init function and free it
again in module_exit.While we are at it simplify a few functions in regard to the
ap configuration data.Reviewed-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
ap_test_queue, ap_query_facilities, __ap_query_functions all use
the same PQAP(TAPQ) command. Consolidate the three into a single
ap_test_queue function that returns the AP status and the 64-bit
result. The exception table entry for PQAP(TAPQ) can be avoided
if the T bit for the APFT facility is set only if test_facility(15)
indicated that the facility is present.Integrate ap_query_function into ap_query queue to avoid calling
PQAP(TAPQ) twice.Reviewed-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky -
The sclp console and tty code currently uses several message text
objects in a single message event to print several lines with one
SCCB. This causes the output of these lines to be fused into a
block which is noticeable when selecting text in the operating system
message panel.Instead use several message events with a single message text object
each to print every line on its own. This changes the SCCB layout
fromstruct sccb_header
struct evbuf_header
struct mdb_header
struct go
struct mto
...
struct mtoto
struct sccb_header
struct evbuf_header
struct mdb_header
struct go
struct mto
...
struct evbuf_header
struct mdb_header
struct go
struct mtoReviewed-by: Peter Oberparleiter
Signed-off-by: Martin Schwidefsky -
Introduce /sys/debug/kernel/diag_stat with a statistic how many diagnose
calls have been done by each CPU in the system.Signed-off-by: Martin Schwidefsky
-
We often need to correlate an 8 bit path mask with the position
in a channel path array. Introduce and use pathmask_to_pos for
that task.Reviewed-by: Peter Oberparleiter
Signed-off-by: Sebastian Ott
Signed-off-by: Martin Schwidefsky -
During resume from hibernate we already reenable measurement block
updates on a per device basis. In addition to that we also need to
activate channel measurement globally using the set channel monitor
instruction.Signed-off-by: Sebastian Ott
Reviewed-by: Martin Schwidefsky
Reviewed-by: Cornelia Huck
Signed-off-by: Martin Schwidefsky -
Extended measurement blocks need to be 64 byte aligned. To achieve that
128 bytes for each measurement block are allocated and an align callback
returns a 64 byte aligned address inside this area.Replace this code with kmem_cache allocations.
Signed-off-by: Sebastian Ott
Reviewed-by: Martin Schwidefsky
Reviewed-by: Cornelia Huck
Signed-off-by: Martin Schwidefsky -
The measurement block for the extended measurement data is not freed when
switching off per device measurement. Free the measurement block after HW
stopped accessing it.Signed-off-by: Sebastian Ott
Reviewed-by: Martin Schwidefsky
Reviewed-by: Cornelia Huck
Signed-off-by: Martin Schwidefsky -
During allocation of extended measurement blocks we check if the device is
already active for channel measurement and add the device to a list of
devices with active channel measurement. The check is done under ccwlock
protection and the list modification is guarded by a different lock.To guarantee that both states are in sync make sure that both locks
are held during the allocation process (like it's already done for the
"normal" measurement block allocation).Signed-off-by: Sebastian Ott
Reviewed-by: Martin Schwidefsky
Signed-off-by: Martin Schwidefsky -
Devices with active channel measurement are included in a list. When a
device is removed without deactivating channel measurement first the
list_head is freed but still used. Fix this by making sure that
channel measurement is deactivated during device deregistration.For devices that we deregister because they are no longer accessible
deactivating channel measurement will fail. In this case we can report
success because the FW will no longer access the measurement block.In addition to these steps keep an extra device reference while
channel measurement is active.Signed-off-by: Sebastian Ott
Reviewed-by: Cornelia Huck
Signed-off-by: Martin Schwidefsky -
Hold the device_lock during [de]activation of the channel measurement
block to synchronize concurrent usage of these functions.Signed-off-by: Sebastian Ott
Reviewed-by: Martin Schwidefsky
Signed-off-by: Martin Schwidefsky -
Ensure that we hold the ccwlock when accessing private data. Return errors
that occur during measurement enabling to userspace. Apply some cleanups
while at it.Signed-off-by: Sebastian Ott
Reviewed-by: Cornelia Huck
Signed-off-by: Martin Schwidefsky -
We were able to reduce the CPU overhead of big paging scenarios
when announcing our paging disks as non-rotational.
Almost all dasd devices are implemented in storage servers with
cache, raid, striping and lots of magic. There is no point in
optimizing the disk schedulers and swap code for a single platter
moving arm rotational disks. Given the complexity of the setup
and the fact that this change is mostly to disable the additional
overhead in swap code, lets keep the other functionality unchanged
and do not disable the this device as entropy source - unlike other
non-rotational devices.Suggested-by: Christian Ehrhardt
Reviewed-by: David Hildenbrand
Signed-off-by: Christian Borntraeger
Signed-off-by: Martin Schwidefsky -
In the past only even modulus sizes were allowed for RSA keys in
CRT format. This restriction was based on limited RSA key generation
on older crypto adapters that provides only even modulus sizes. This
restriction is not valid any more.Revoke restrictions that crypto requests can be serviced with odd
RSA modulus length in CRT format.Signed-off-by: Ingo Tuchscherer
Signed-off-by: Martin Schwidefsky
07 Oct, 2015
3 commits
-
If HiperSockets Completion Queueing is enabled, qdio always
issues a warning, since the condition is always met.
This patch fixes the condition in WARN_ON_ONCE that was always
true.Signed-off-by: Eugene Crosser
Signed-off-by: Ursula Braun
Signed-off-by: David S. Miller -
In layer2 mode of the qeth driver, MAC address lists
from struct net_device require mapping to the OSA-card.
The existing implementation is inefficient for lists with
more than several MAC addresses, since for every
ndo_set_rx_mode callback it removes all MAC addresses first,
and then registers the current MAC address list.
This patch changes implementation of ndo_set_rx_mode callback
in qeth, only performing hardware registration/removal for
new/deleted addresses. To shorten lookup of MAC addresses
registered addresses are kept in a hashtable instead of a
linear list.Signed-off-by: Lakhvich Dmitriy
Signed-off-by: Ursula Braun
Reviewed-by: Eugene Crosser
Reviewed-by: Thomas Richter
Tested-by: Christian Borntraeger
Signed-off-by: David S. Miller -
Add support for GRO (generic receive offload) in the layer 2
part of device driver qeth. This results in a performance
improvement when GRO and RX is turned on.Signed-off-by: Thomas Richter
Signed-off-by: Ursula Braun
Signed-off-by: David S. Miller
