The kernel's implementation of notifier chains is unsafe. There is no
protection against entries being added to or removed from a chain while the
chain is in use. The issues were discussed in this thread:
http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2
We noticed that notifier chains in the kernel fall into two basic usage
classes:
"Blocking" chains are always called from a process context
and the callout routines are allowed to sleep;
"Atomic" chains can be called from an atomic context and
the callout routines are not allowed to sleep.
We decided to codify this distinction and make it part of the API. Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name). New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain. The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.
With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed. For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections. (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)
There are some limitations, which should not be too hard to live with. For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem. Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain. (This did happen in a couple of places and the code
had to be changed to avoid it.)
Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization. Instead we use RCU. The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.
Here is the list of chains that we adjusted and their classifications. None
of them use the raw API, so for the moment it is only a placeholder.
ATOMIC CHAINS
-------------
arch/i386/kernel/traps.c: i386die_chain
arch/ia64/kernel/traps.c: ia64die_chain
arch/powerpc/kernel/traps.c: powerpc_die_chain
arch/sparc64/kernel/traps.c: sparc64die_chain
arch/x86_64/kernel/traps.c: die_chain
drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list
kernel/panic.c: panic_notifier_list
kernel/profile.c: task_free_notifier
net/bluetooth/hci_core.c: hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain
net/ipv6/addrconf.c: inet6addr_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain
net/netlink/af_netlink.c: netlink_chain
BLOCKING CHAINS
---------------
arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain
arch/s390/kernel/process.c: idle_chain
arch/x86_64/kernel/process.c idle_notifier
drivers/base/memory.c: memory_chain
drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list
drivers/macintosh/adb.c: adb_client_list
drivers/macintosh/via-pmu.c sleep_notifier_list
drivers/macintosh/via-pmu68k.c sleep_notifier_list
drivers/macintosh/windfarm_core.c wf_client_list
drivers/usb/core/notify.c usb_notifier_list
drivers/video/fbmem.c fb_notifier_list
kernel/cpu.c cpu_chain
kernel/module.c module_notify_list
kernel/profile.c munmap_notifier
kernel/profile.c task_exit_notifier
kernel/sys.c reboot_notifier_list
net/core/dev.c netdev_chain
net/decnet/dn_dev.c: dnaddr_chain
net/ipv4/devinet.c: inetaddr_chain
It's possible that some of these classifications are wrong. If they are,
please let us know or submit a patch to fix them. Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)
The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.
[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add full driver model support for the IPMI driver. It links in the proper
bus and device support.
It adds an "ipmi" driver interface that has each BMC discovered by the
driver (as a device). These BMCs appear in the devices/platform directory.
If there are multiple interfaces to the same BMC, the driver should
discover this and will only have one BMC entry. The BMC entry will have
pointers to each interface device that connects to it.
The device information (statistics and config information) has not yet been
ported over to the driver model from proc, that will come later.
This work was based on work by Yani Ioannou. I basically rewrote it using
that code as a guide, but he still deserves credit :).
[bunk@stusta.de: make ipmi_find_bmc_guid() static]
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Yani Ioannou <yani.ioannou@gmail.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Modify the PCI hanling code for the IPMI driver to use the new method of
tables and registering, and adds more generic PCI handling for IPMI.
Unfortunately, this required a rather large rework of the way the driver
did detection so it would be more event-driven.
[bunk@stusta.de: make a struct static]
Signed-off-by: Corey Minyard <minyard@acm.org>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
On mips:
drivers/char/ipmi/ipmi_si_intf.c:1274: error: conflicting types for 'mem_inb'
include/asm/io.h:436: error: previous definition of 'mem_inb' was here
Don't look at line 436 unless you really know what you're doing.
Move those static functions out of more or less generic namespace.
Signed-off-by: Alexey "## should be banned" Dobriyan <adobriyan@gmail.com>
Acked-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
At the 2.6.12 timeframe ipmi_si_intf.c was patched to provide default
register spacings in try_init_acpi() if the register spacing was set to
zero, similar to code in other routines.
Unfortunately, another patch was simultaneously added that exits early from
try_init_acpi() if the register spacings are set to zero, circumventing the
new defaults. This patch removes the early exit code and some incorrect
comments that aren't present in other common code snippets.
Signed-off-by: Rocky Craig <rocky.craig@hp.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With Andi Kleen's x86_64 patch to use DMI, and my ia64 to use DMI, there is
now a new CONFIG_DMI option which takes the place of CONFIG_X86 to denote
the availability of the DMI functions. Make the IPMI driver use CONFIG_DMI
instead.
Tested on ia64 2.6.15 kernel plus the previous patch, on a Dell PowerEdge
7250 Itanium2 server, and it now autodetects the IPMI KCS driver as
expected.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Acked-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
While doing some testing I discovered that if the BIOS on a board does not
properly setup the DMI information it leads to a panic in the IPMI code.
The panic is due to dereferencing a pointer which is not initialized. The
pointer is initialized in port_setup() and/or mem_setup() and used in
init_one_smi() and cleanup_one_si(), however if either port_setup() or
mem_setup() return ENODEV the pointer does not get initialized.
Signed-off-by: Paolo Galtieri <pgaltieri@mvista.com>
Acked-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
On IPMI systems with BT interfaces, we don't start the kernel thread, so
smi_info->thread is NULL. Test for NULL when stopping the thread, because
kthread_stop() doesn't, and an oops ensues otherwise.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Acked-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Convert ipmi driver thread to kthread API, only sleep when interface is
idle.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Cc: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We must poll for responses to commands when interrupts aren't in use. The
default poll interval is based on using a kernel timer, which varies with HZ.
For character-based interfaces like KCS and SMIC though, that can be way too
slow (>15 minutes to flash a new firmware with KCS, >20 seconds to retrieve
the sensor list).
This creates a low-priority kernel thread to poll more often. If the state
machine is idle, so is the kernel thread. But if there's an active command,
it polls quite rapidly. This decrease a firmware flash time from 15 minutes
to 1.5 minutes, and the sensor list time to 4.5 seconds, on a Dell PowerEdge
x8x system.
The timer-based polling remains, to ensure some amount of responsiveness even
under high user process CPU load.
Checking for a stopped timer at rmmod now uses atomics and del_timer_sync() to
ensure safe stoppage.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
BMCs can get into ERROR0 state while flashing new firmware, particularly while
the BMC is erasing the next flash block, which may take a just under 2 seconds
on a Dell PowerEdge 2800 (1.75 seconds typical), during which time the
single-threaded firmware may not be able to process new commands. In
particular, clearing OBF may not take effect immediately.
We want it to delay in ERROR0 after clearing OBF a bit waiting for OBF to
actually be clear before proceeding.
This introduces a new return value from the LLDD's event loop,
SI_SM_CALL_WITH_TICK_DELAY. This means the calling thread/timer should
schedule_timeout() at least 1 tick, rather than busy-wait. This is a longer
delay than SI_SM_CALL_WITH_DELAY, which is typically a 250us busy-wait.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Some commands, on some system BMCs, don't respond at at all. This is seen on
Dell PowerEdge x6xx and x7xx systems with IPMI 1.0 BT controllers when a "Get
SDR" command is issued, with a length field of 0x3A, which happens to be the
length of about SDR entries. If another length is passed, this command
succeeds.
This patch adds general infrastructure for receiving commands before they're
passed down to the low-level drivers, such that they can be completed
immediately, or modified, prior to being sent to ->start_transaction().
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Make SMIC driver ignore EVT_AVAIL and SMS_ATN bits in flags register, as
they're used by systems management interrupts, not the host OS.
Make the OEM0 Data Available handler work for pre-IPMI 1.5 systems from Dell
too.
Without these two fixes, PowerEdge 2650 and other similar systems with SMIC
may hang a process (modprobe or anything using /dev/ipmi0).
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
A number of small changes for the various system interface drivers,
consolidated from a number of patches from Matt Domsch.
Clear B2H_ATN and drain the BMC message buffer on command timeout. This
prevents further commands from failing after a timeout.
Add bt_debug and smic_debug module parameters, expose them in sysfs. This
lets you enable and disable debugging messages at runtime.
Unsigned jiffies math in ipmi_si_intf.c causes a too-large value to be passed
to ->event() after jiffies wrap-around. The BT driver had caught this, but
didn't know how to fix it. Now all calls to ->event() use a sane value for
time.
Increase timeout for commands handed to the BT driver from 2 seconds to 5
seconds. This is necessary particularly when the previous command was a
"Clear SEL", as that command completes, yet the BMC isn't really ready to
handle another command yet.
Silence BT debugging messages which were being printed on the console.
Increase SMIC timeout form 1/10s to 2s. This is needed on Dell PowerEdge 2650
and PowerEdge 750 with ERA/O cards to allow commands to complete without
timing out.
Adds kcs_debug module param, to match behavior of BT and SMIC. This also
prevents messages from being sent to the console unless explicitly requested.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Clean up various style issues in the IPMI driver. Should be no functional
changes.
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This adds MODULE_VERSION, MODULE_DESCRIPTION, and MODULE_AUTHOR tags to the
IPMI driver modules. Also changes the MODULE_VERSION to remove the
prepended 'v' on each value, consistent with the module versioning policy.
This patch also removes all the version information from everything except
the ipmi_msghandler module.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The ipmi driver does not have a way to handle firmware-generated events
which have the OEM[012] Data Available flags set. In such a case, the
SMS_ATN bit may never get cleared by firmware, leaving the driver looping
infinitely but never able to make any progress.
This patch first simplifies storage and use of the data returned from an
IPMI Get Device ID command.
It then creates a new per-OEM handler hook, which should know how to handle
events with the OEM[012] Data Available flags set. It then uses this to
implement a workaround for IPMI 1.5-capable Dell PowerEdge servers which
are susceptable to setting the OEM[012] Data Available flags when the
driver can't handle it.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Fix some problems with the high-res timer support.
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
`gcc -W' likes to complain if the static keyword is not at the beginning of
the declaration. This patch fixes all remaining occurrences of "inline
static" up with "static inline" in the entire kernel tree (140 occurrences in
47 files).
While making this change I came across a few lines with trailing whitespace
that I also fixed up, I have also added or removed a blank line or two here
and there, but there are no functional changes in the patch.
Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
annotated, a bunch of direct dereferencing replaced with readb().
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Enable interrupts for a BT interface. There is a specific register that
needs to be set up to enable interrupts that also must be modified to clear
the irq.
Also, don't reset the BMC on a BT interface. That's probably not a good
idea as the BMC may be performing other important functions and a reset
should only be a last resort. Also, that register is also used to
enable/disable interrupts to the BT; modifying it may screw up the
interrupts.
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If the ACPI register bit width is zero (an invalid value) assume it is the
default spacing. This avoids some coredumps on invalid data and makes some
systems work that have broken ACPI data.
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Ignore the bottom bit of the base address from the DMI data. It is
supposed to be set to 1 if it is I/O space. Few systems do this, but this
enables the ones that do set it to work properly.
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch changes calls to synchronize_kernel(), deprecated in the earlier
"Deprecate synchronize_kernel, GPL replacement" patch to instead call the new
synchronize_rcu() and synchronize_sched() APIs.
Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
