Jack Ren and Eric Miao tracked down the following long standing
problem in the NOHZ code:
scheduler switch to idle task
enable interrupts
Window starts here
----> interrupt happens (does not set NEED_RESCHED)
irq_exit() stops the tick
----> interrupt happens (does set NEED_RESCHED)
return from schedule()
cpu_idle(): preempt_disable();
Window ends here
The interrupts can happen at any point inside the race window. The
first interrupt stops the tick, the second one causes the scheduler to
rerun and switch away from idle again and we end up with the tick
disabled.
The fact that it needs two interrupts where the first one does not set
NEED_RESCHED and the second one does made the bug obscure and extremly
hard to reproduce and analyse. Kudos to Jack and Eric.
Solution: Limit the NOHZ functionality to the idle loop to make sure
that we can not run into such a situation ever again.
cpu_idle()
{
preempt_disable();
while(1) {
tick_nohz_stop_sched_tick(1); <- tell NOHZ code that we
are in the idle loop
while (!need_resched())
halt();
tick_nohz_restart_sched_tick(); <- disables NOHZ mode
preempt_enable_no_resched();
schedule();
preempt_disable();
}
}
In hindsight we should have done this forever, but ...
/me grabs a large brown paperbag.
Debugged-by: Jack Ren <jack.ren@marvell.com>,
Debugged-by: eric miao <eric.y.miao@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Possible RETS Register Corruption when Subroutine Is under 5 Cycles in Duration
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Enable: PM_SUSPEND_MEM -> Blackfin Hibernate to SDRAM
This feature requires a special bootloader (u-boot)
supporting return from hibernate.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
use kernel command line mem and max_mem bootargs to limit
availabe memory instead.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
--
WARNING: vmlinux.o(.text+0x721a): Section mismatch in reference from the function ___fill_code_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_code_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_code_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x7238): Section mismatch in reference from the function ___fill_code_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_code_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_code_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x7250): Section mismatch in reference from the function ___fill_code_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_code_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_code_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x7264): Section mismatch in reference from the function ___fill_code_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_code_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_code_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x72a2): Section mismatch in reference from the function ___fill_data_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_data_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_data_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x72bc): Section mismatch in reference from the function ___fill_data_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_data_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_data_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x72d4): Section mismatch in reference from the function ___fill_data_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_data_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_data_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
WARNING: vmlinux.o(.text+0x72e8): Section mismatch in reference from the function ___fill_data_cplbtab() to the function .init.text:_fill_cplbtab()
The function ___fill_data_cplbtab() references
the function __init _fill_cplbtab().
This is often because ___fill_data_cplbtab lacks a __init
annotation or the annotation of _fill_cplbtab is wrong.
--
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Initialize the lock of bad_irq_desc properly.
The content of irq_desc array is replaced by bad_irq_desc in blackfin
arch irqchip init code. So, do it properly as common irq init code.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
[PATCH] return to old errno choice in mkdir() et.al.
[Patch] fs/binfmt_elf.c: fix wrong return values
[PATCH] get rid of leak in compat_execve()
[Patch] fs/binfmt_elf.c: fix a wrong free
[PATCH] avoid multiplication overflows and signedness issues for max_fds
[PATCH] dup_fd() part 4 - race fix
[PATCH] dup_fd() - part 3
[PATCH] dup_fd() part 2
[PATCH] dup_fd() fixes, part 1
[PATCH] take init_files to fs/file.c
so that we always send the same signal and we handle the NULL ptr condition properly
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
IMDMA does not operate to full speed for 600MHz and higher devices
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Remove module will not free L1 memory used which caused by
memory access after free. This patch fixes it.
Signed-off-by: Meihui Fan <mhfan@hhcn.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
When delivering a signal, disable single stepping but call
ptrace_notify if it was enabled before. The idea was taken
from the x86 port.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Fix some really ancient code that was correct only for the m68k port.
Delete unused (i.e. copied from m68k) entries in asm-offsets.c.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
The following cleanup patch:
add __user markings to a few userspace system functions
mysteriously added a "&" operator that doesn't belong in there, breaking the
atomic sections code.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
This replaces the duplicated arch-specific versions of "sys_pipe()" with
one unified implementation. This removes almost 250 lines of duplicated
code.
It's marked __weak, so that *if* an architecture wants to override the
default implementation it can do so by simply having its own replacement
version, since many architectures use alternate calling conventions for
the 'pipe()' system call for legacy reasons (ie traditional UNIX
implementations often return the two file descriptors in registers)
I still haven't changed the cris version even though Linus says the BKL
isn't needed. The arch maintainer can easily do it if there are really
no obstacles.
Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove the circular buffering mechanism for exceptions. Instead, point RETX
at a safe location from which to fetch three NOPs.
This safe location is now in the fixed code area, and also used for certain
anomaly workarounds, to ensure that user space can find a valid ICPLB when
things are built with CONFIG_MPU.
Also, save I/DCPLB_FAULT_ADDRESS when lowering to level 5, since the hardware
reg is valid only at exception level.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
- add platform device resources in board files
- add new bfin_sir.h to each machines
Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
The irq setup code no longer calls gpio request and free.
This patch also changes the default gpio_free behavior on Blackfin.
A freed GPIO keeps it's last state, and is not defaulted back to
an input. This is also what all other architectures do.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
detect the memory available in the system on the fly by default
rather than forcing people to set this manually in the kconfig
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
http://blackfin.uclinux.org/gf/project/uclinux-dist/tracker/?action=TrackerItemEdit&tracker_item_id=3978
Section data_l1_cacheline_aligned should be defined in
link script of kernel, when L1 data sram bank A is not available.
In bf536 with all data cache is enabled, there is no L1 data sram.
Current link script won't define section data_l1.cacheline_aligned in
this case. But, if user select put cacheline_aligned data into l1 sram
in kernel menuconfig, these data will be dropped and access to these
data will trigger data CPLB exception.
Do panic in l1 relocation code as well.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
- allow bootrom to be readable from supervisor mode
- delete unused local variable "addr"
- punt unused local defines of cplbinfo.c
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
The DMA base registers are available in a global named "base_addr" for
every Blackfin variant. Give this a more descriptive name, and remove
duplicate tables from some drivers.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
theres no need to declare ram{end,start,base} in the head.S files
when declaring them with the other memory related variables in setup.c
is so much simpler/nicer
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
remove duplicated prototypes for internal cplb structures from
the global blackfin header as nothing else should be accessing these
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Previously, init failed to do anything meaningful;
it turns out that the reason is that FD-PIC has a readonly data
section which can be located in the XIP filesystem, and various address checks
in the kernel reject such addresses for syscall arguments. Hence, init's
execve ("/bin/sh", ...)
failed with error code EFAULT.
There's room for improvement here: in case people want to have filesystems
on flash rather than in main memory, _access_ok should be modified to
allow this.
This bug fix is also dedicated to Michael Hennerich.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
The function flush_switched_dcplbs was clearing the CPLB entries covering
the process permission bitmasks. This means that the sequence
flush_switched_dcplbs ();
set_mask_dcplbs(mm->context.page_rwx_mask);
has a problem: if kernel code (such as an interrupt) causes a CPLB miss before
set_mask_dcplbs completes, the CPLB handler function causes a double fault,
with an instantaneous reboot.
This bug fix is dedicated to Michael Hennerich, the only person in the world
capable of providing working JTAG hardware.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
This is a rather old performance improvement for the signal handling
code, which was originally only committed on the 2007R1 branch as a
workaround for what we suspected to be a hardware bug.
There's no point in constructing a sigreturn stub on the stack and
flushing caches; we can just make signal handlers return to a known
location in the fixed code area.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
relocate MAX_SWITCH_{D,I}_CPLBS from the header to the file
where it actually gets used. this way when we change
CONFIG_MEM_SIZE in our kconfig, we only rebuild one or two files
rather than a whole bunch that implicitly include cplb.h.
this will also remove the ability to clear the swapcount on
the fly, but i really dont think that functionality is important.
ultimate goal is for CONFIG_MEM_SIZE to go away and calculate
this value on the fly based on what u-boot programmed for us.
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
grab locks when not atomic - this fixes the issues
sometimes seen when using magic sysrq.
Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Sometimes when we crash, current is not valid, (has been written
over), so the existing code causes a invalid read during exception
context - which is a unrecoverable double fault. This fixes this.
Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Since
r3658 | vapier | 2007-09-12 16:26:11 +0200 (Wed, 12 Sep 2007) | 1 line
add more common defines for output sections
we've had a new line, NOTES, in our linker script, which causes upstream
binutils to complain about "missing phdr". Currently the only other arch
that uses NOTES is i386, and the patch which added it also added
PHDRS {
text PT_LOAD FLAGS(5); /* R_E */
data PT_LOAD FLAGS(7); /* RWE */
note PT_NOTE FLAGS(0); /* ___ */
}
and a few other modifications to use ":text" and ":data" to the linker
script.
It seems that we don't need NOTES at all, so just remove it.
Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Semaphores are no longer performance-critical, so a generic C
implementation is better for maintainability, debuggability and
extensibility. Thanks to Peter Zijlstra for fixing the lockdep
warning. Thanks to Harvey Harrison for pointing out that the
unlikely() was unnecessary.
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
d_path() is used on a <dentry,vfsmount> pair. Lets use a struct path to
reflect this.
[akpm@linux-foundation.org: fix build in mm/memory.c]
Signed-off-by: Jan Blunck <jblunck@suse.de>
Acked-by: Bryan Wu <bryan.wu@analog.com>
Acked-by: Christoph Hellwig <hch@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Neil Brown <neilb@suse.de>
Cc: Michael Halcrow <mhalcrow@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit d3d74453c3 ("hrtimer: fixup the
HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback") broke several archs, and since
only Russell bothered to merge the fix, and Greg to ACK his arch, I'm
sending this for merger.
I have confirmation that the Alpha bit results in a booting kernel.
That leaves: blackfin, frv, sh and sparc untested.
The deadlock in question was found by Russell:
IRQ handle
-> timer_tick() - xtime seqlock held for write
-> update_process_times()
-> run_local_timers()
-> hrtimer_run_queues()
-> hrtimer_get_softirq_time() - tries to get a read lock
Now, Thomas assures me the fix is trivial, only do_timer() needs to be
done under the xtime_lock, and update_process_times() can savely be
removed from under it.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Greg Ungerer <gerg@uclinux.org>
CC: Richard Henderson <rth@twiddle.net>
CC: Bryan Wu <bryan.wu@analog.com>
CC: David Howells <dhowells@redhat.com>
CC: Paul Mundt <lethal@linux-sh.org>
CC: William Irwin <wli@holomorphy.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move the init sections to the end of memory so that after they
are free, run time memory is all continugous - this should help decrease
memory fragementation.
When doing this, we also pack some of the other sections a little closer
together, to make sure we don't waste memory. To make this happen,
we need to rename the .data.init_task section to .init_task.data, so
it doesn't get picked up by the linker script glob.
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
This fixes a bug (zero pointer access) only seen on BF561, during USB
Mass Storage/SCSI Host initialization.
It appears to be related to registering a none existing CPU
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
- Add support for irq_wake on system and gpio interrupts
- Remove outdated kernel options
- Add option to select default PM mode
- Fix various places where SIC_IWRx was only handled partially
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
This patchset adds a flags variable to reserve_bootmem() and uses the
BOOTMEM_EXCLUSIVE flag in crashkernel reservation code to detect collisions
between crashkernel area and already used memory.
This patch:
Change the reserve_bootmem() function to accept a new flag BOOTMEM_EXCLUSIVE.
If that flag is set, the function returns with -EBUSY if the memory already
has been reserved in the past. This is to avoid conflicts.
Because that code runs before SMP initialisation, there's no race condition
inside reserve_bootmem_core().
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix powerpc build]
Signed-off-by: Bernhard Walle <bwalle@suse.de>
Cc: <linux-arch@vger.kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch consolidate all definitions of .init.text, .init.data
and .exit.text, .exit.data section definitions in
the generic vmlinux.lds.h.
This is a preparational patch - alone it does not buy
us much good.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Some of the information in kernel boot message is not reasonable.
http://blackfin.uclinux.org/gf/project/uclinux-dist/tracker/?action=TrackerItemEdit&tracker_item_id=3846
- use _rambase as the start of kernel image.
kernel is in the region [_rambase, _ramstart]
- count in pages in per-cpu-page list as available memory
- reserved memory now include: [0 - 4K] for bad pointer catching,
memory reserved for abnormaly 05000263, memory reserved by kernel itself.
Signed-off-by: Yi Li <yi.li@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
writes to I/DMEM_CONTROL must be followed by SSYNC
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
- Add anomaly workaround for bfin_gpio_reset_spi0_ssel1
- Fix style
- Update copyright
- Remove BUG_ON checks for functions intended to be used only by arch
support. GPIO users should only access using the generic GPIO API
- Make all GPIO identifier unsigned int
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Don't oops_in_progress if single step is comming from the
kernel, which happens if a single step occurs after a exception cause.
This fixes up the remaining issues in the toolchain bug.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Make sure the SYSTEM reset completes before we issue the CORE reset
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
early serial init also utilizes the peripheral request api - however
at this point bfin_gpio_init didn't allocate memory for the labels.
So we always have two zombies (allocated pin functions without labels)
This happens before the initcalls - We now allocate memory statically.
Define MAX_RESOURCES individually for each cpu.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
if it does get re-added, it needs to be in the boards directory,
not common code ... or it needs a re-implementation
Signed-off-by: Mike Frysinger <michael.frysinger@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Clean up dump_bfin_mem so that it will display
content from the kernel, as well as l1 instruction, when deferred
HW errors happen, print out the last frame info if it makes sense.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
http://blackfin.uclinux.org/gf/project/uclinux-dist/tracker/?action=TrackerItemEdit&tracker_item_id=3719
When the CPLBs get a miss, we do:
- find a victim in the HW table
- remove the victim
- find the replacement in the software table
- put it into the HW table.
If we can't find a replacement in the software table, we accidently
leave a duplicate in the HW table. This patch ensures that duplicate
is marked as not valid.
What we should do is find the replacement in the software table, before
we find a victim in the HW table - but its too late in the release cycle
to do that much restructuring of this code.
Rather that duplicate code, connect Hardware Errors (irq5) into trap_c,
so user space processes get killed properly.
The rest of irq_panic() can be moved into traps.c (later)
There is still a small corner case that causes problems when a
pheriperal interrupt goes off a single cycle before a user space
hardware error. This causes a kernel panic, rather than the user
space process being killed.
But, this checkin makes things work in 99.9% of the cases, and is a vast
improvement from what is there today (which fails 100% of the time).
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
We need to send signals with the proper PC, or gdb gets
confused, and lots of tests fail. This should fix that.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
- remove duplicated defines for the BF561
- generalize L2 support (so that it works for BF54x) and mark it executable
- add support for reading/executing the Boot ROM sections
(since it has data/functions we may need at runtime)
- and fixup names for each map
Signed-off-by: Mike Frysinger <michael.frysinger@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
- no need to declare their sizes in the common header
- no need to tack on the section attribute as only the definition matters, not references
Signed-off-by: Mike Frysinger <michael.frysinger@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Revert this patch:
move the init sections to the end of memory, so that after they
are free, run time memory is all continugous - this should help decrease
memory fragementation. When doing this, we also pack some of the other
sections a little closer together, to make sure we don't waste memory.
To make this happen, we need to rename the .data.init_task section to
.init_task.data, so it doesn't get picked up by the linker script glob.
Since it causes the kernel not to boot up with mtd filesystems.
Signed-off-by: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
/*
* CPUs often take a performance hit when accessing unaligned memory
* locations. The actual performance hit varies, it can be small if the
* hardware handles it or large if we have to take an exception and fix
* it
* in software.
*
* Since an ethernet header is 14 bytes network drivers often end up
* with
* the IP header at an unaligned offset. The IP header can be aligned by
* shifting the start of the packet by 2 bytes. Drivers should do this
* with:
*
* skb_reserve(NET_IP_ALIGN);
*
* The downside to this alignment of the IP header is that the DMA is
* now
* unaligned. On some architectures the cost of an unaligned DMA is high
* and this cost outweighs the gains made by aligning the IP header.
*
* Since this trade off varies between architectures, we allow
* NET_IP_ALIGN
* to be overridden.
*/
This new function insl_16 allows to read form 32-bit IO and writes to
16-bit aligned memory. This is useful in above described scenario -
In particular with the AXIS AX88180 Gigabit Ethernet MAC.
Once the device is in 32-bit mode, reads from the RX FIFO always
decrements 4bytes.
While on the other side the destination address in SDRAM is always
16-bit aligned.
If we use skb_reserve(0) the receive buffer is 32-bit aligned but later
we hit a unaligned exception in the IP code.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
If you need a 64 bit divide in the kernel, use asm/div64.h.
Revert the addition of udivdi3.
Cc: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: Mike Frysinger <michael.frysinger@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
The only user of get_wchan I was able to find is the proc fs - and proc
can't be built modular.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
The only user is the a.out support.
It was therefore removed prior to the blackfin merge from all
architectures not supporting a.out.
Currently, Blackfin doesn't suppport a.out.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
move the init sections to the end of memory, so that after they
are free, run time memory is all continugous - this should help decrease
memory fragementation. When doing this, we also pack some of the other
sections a little closer together, to make sure we don't waste memory.
To make this happen, we need to rename the .data.init_task section to
.init_task.data, so it doesn't get picked up by the linker script glob.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
We currently do not. Also make it easier to handle cplb violations - in traps.c
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
This is fixes a problem where we could jump to the wrong address. By
doing a "p0 = reti; jump (p0)". If a different, higher level interrupt
came in, just before, rather than returning to the calling function, we
would return to a random place in the kernel.
This very elegant fix from Bernd grabs the return location off the
stack, and places it into P0, so when we do a return, it goes to the
correct place.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
- move the CONFIG_KGDB into one block, for easier reading
- remove printk from printk_address, and pass around buffers. Also
print out the labels when decoding CPLB errors, so you know exactly
where the error was.
- Do not use fixed addresses, becuase people do not know where they come from.
- Turn the printing level down on the dump, so if you don't want,
only the signal prints out - just like on other archs. If a kernel/interrupt
crashes, it should dump everything all the time
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
This fixes two things:
- stop calling write_lock_irq/write_unlock_irq which can turn modify
irq levels
- don't calling mmput when handing exceptions - since this might_sleep,
which does a rti, and leaves us in kernel space (irq15, rather
than irq5).
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/cooloney/blackfin-2.6:
Blackfin arch: use KBUILD_CFLAGS and KBUILD_AFLAGS in Makefile
Blackfin arch: Javier Herrer writes: fix building when icache and dcache is disabled
We must balance calls to get_task_mm with corresponding mmput calls, otherwise
refcounting is screwed up and mms don't get freed when their task exits.
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Fix up /proc/cpuinfo so it is like everyone else, and gets
parsed by various applications properly. Still needs some tweaking on
parts without full L1 sram, like 532, 531, so it doesn't print out L1
bank info that doesn't exist.
Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Identical handlers of PTRACE_DETACH go into ptrace_request().
Not touching compat code.
Not touching archs that don't call ptrace_request.
Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Acked-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix a problem reported in the forums - libstdc++ can call writev with an
iovec containing { NULL, 0 }, which works fine on i686-linux, but fails on
Blackfin. Fixed by allowing size 0 transfers to/from userspace regardless
of the address.
Signed-off-by: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
- flush/inv the correct range
- dmacopy test failed when policy is write_back - invalidate before dma
http://blackfin.uclinux.org/gf/project/uclinux-dist/tracker/?action=TrackerItemEdit&tracker_item_id=3367
It's the cache invalidate what is causing the issue.
There is no invalidate only instruction it's always: FLUSHINV
So when we "invalidate" after the DMA we might (do) overwrite freshly
dma'ed data by dirty Cache WB content.
Fixed by moving the "invalidate" at the beginning of dma_memcpy.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
