Improve the scalability of the IDA by using the per-IDA xa_lock rather
than the global simple_ida_lock. IDAs are not typically used in
performance-sensitive locations, but since we have this lock anyway, we
can use it. It is also a step towards converting the IDA from the radix
tree to the XArray.
[akpm@linux-foundation.org: idr.c needs xarray.h]
Link: http://lkml.kernel.org/r/20180331125332.GF13332@bombadil.infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Khalid reported that the kernel selftests are currently failing:
selftests: test_bpf.sh
========================================
test_bpf: [FAIL]
not ok 1..8 selftests: test_bpf.sh [FAIL]
He bisected it to 6ce711f275 ("idr: Make
1-based IDRs more efficient").
The root cause is doing a signed comparison in idr_alloc_u32() instead
of an unsigned comparison. I went looking for any similar problems and
found a couple (which would each result in the failure to warn in two
situations that aren't supposed to happen).
I knocked up a few test-cases to prove that I was right and added them
to the test-suite.
Reported-by: Khalid Aziz <khalid.aziz@oracle.com>
Tested-by: Khalid Aziz <khalid.aziz@oracle.com>
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
As far as I can tell, the only place the per-cpu ida_bitmap is populated
is in ida_pre_get. The pre-allocated element is stolen in two places in
ida_get_new_above, in both cases immediately followed by a memset(0).
Since ida_get_new_above is called with locks held, do the zeroing in
ida_pre_get, or rather let kmalloc() do it. Also, apparently gcc
generates ~44 bytes of code to do a memset(, 0, 128):
$ scripts/bloat-o-meter vmlinux.{0,1}
add/remove: 0/0 grow/shrink: 2/1 up/down: 5/-88 (-83)
Function old new delta
ida_pre_get 115 119 +4
vermagic 27 28 +1
ida_get_new_above 715 627 -88
Link: http://lkml.kernel.org/r/20180108225634.15340-1-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Acked-by: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Eric Biggers <ebiggers@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
About 20% of the IDR users in the kernel want the allocated IDs to start
at 1. The implementation currently searches all the way down the left
hand side of the tree, finds no free ID other than ID 0, walks all the
way back up, and then all the way down again. This patch 'rebases' the
ID so we fill the entire radix tree, rather than leave a gap at 0.
Chris Wilson says: "I did the quick hack of allocating index 0 of the
idr and that eradicated idr_get_free() from being at the top of the
profiles for the many-object stress tests. This improvement will be
much appreciated."
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Now that the IDR can be used to store large IDs, it is possible somebody
might only partially convert their old code and use the iterators which
can only handle IDs up to INT_MAX. It's probably unwise to show them a
truncated ID, so settle for spewing warnings to dmesg, and terminating
the iteration.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Most places in the kernel that we need to distinguish functions by the
type of their arguments, we use '_ul' as a suffix for the unsigned long
variant, not '_ext'. Also add kernel-doc.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
It has no more users, so remove it. Move idr_alloc() back into idr.c,
move the guts of idr_alloc_cmn() into idr_alloc_u32(), remove the
wrappers around idr_get_free_cmn() and rename it to idr_get_free().
While there is now no interface to allocate IDs larger than a u32,
the IDR internals remain ready to handle a larger ID should a need arise.
These changes make it possible to provide the guarantee that, if the
nextid pointer points into the object, the object's ID will be initialised
before a concurrent lookup can find the object.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
All current users of idr_alloc_ext() actually want to allocate a u32
and idr_alloc_u32() fits their needs better.
Like idr_get_next(), it uses a 'nextid' argument which serves as both
a pointer to the start ID and the assigned ID (instead of a separate
minimum and pointer-to-assigned-ID argument). It uses a 'max' argument
rather than 'end' because the semantics that idr_alloc has for 'end'
don't work well for unsigned types.
Since idr_alloc_u32() returns an errno instead of the allocated ID, mark
it as __must_check to help callers use it correctly. Include copious
kernel-doc. Chris Mi <chrism@mellanox.com> has promised to contribute
test-cases for idr_alloc_u32.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Changing idr_replace's 'id' argument to 'unsigned long' works for all
callers. Callers which passed a negative ID now get -ENOENT instead of
-EINVAL. No callers relied on this error value.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
During truncation, the mapping has already been checked for shmem and
dax so it's known that workingset_update_node is required.
This patch avoids the checks on mapping for each page being truncated.
In all other cases, a lookup helper is used to determine if
workingset_update_node() needs to be called. The one danger is that the
API is slightly harder to use as calling workingset_update_node directly
without checking for dax or shmem mappings could lead to surprises.
However, the API rarely needs to be used and hopefully the comment is
enough to give people the hint.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
oneirq-v1r1 pickhelper-v1r1
Min Time 141.00 ( 0.00%) 140.00 ( 0.71%)
1st-qrtle Time 142.00 ( 0.00%) 141.00 ( 0.70%)
2nd-qrtle Time 142.00 ( 0.00%) 142.00 ( 0.00%)
3rd-qrtle Time 143.00 ( 0.00%) 143.00 ( 0.00%)
Max-90% Time 144.00 ( 0.00%) 144.00 ( 0.00%)
Max-95% Time 147.00 ( 0.00%) 145.00 ( 1.36%)
Max-99% Time 195.00 ( 0.00%) 191.00 ( 2.05%)
Max Time 230.00 ( 0.00%) 205.00 ( 10.87%)
Amean Time 144.37 ( 0.00%) 143.82 ( 0.38%)
Stddev Time 10.44 ( 0.00%) 9.00 ( 13.74%)
Coeff Time 7.23 ( 0.00%) 6.26 ( 13.41%)
Best99%Amean Time 143.72 ( 0.00%) 143.34 ( 0.26%)
Best95%Amean Time 142.37 ( 0.00%) 142.00 ( 0.26%)
Best90%Amean Time 142.19 ( 0.00%) 141.85 ( 0.24%)
Best75%Amean Time 141.92 ( 0.00%) 141.58 ( 0.24%)
Best50%Amean Time 141.69 ( 0.00%) 141.31 ( 0.27%)
Best25%Amean Time 141.38 ( 0.00%) 140.97 ( 0.29%)
As you'd expect, the gain is marginal but it can be detected. The
differences in bonnie are all within the noise which is not surprising
given the impact on the microbenchmark.
radix_tree_update_node_t is a callback for some radix operations that
optionally passes in a private field. The only user of the callback is
workingset_update_node and as it no longer requires a mapping, the
private field is removed.
Link: http://lkml.kernel.org/r/20171018075952.10627-3-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
idr_replace() returns the old value on success, not 0.
Link: http://lkml.kernel.org/r/20170918162642.37511-1-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
IDR only supports non-negative IDs. There used to be a 'WARN_ON_ONCE(id <
0)' in idr_replace(), but it was intentionally removed by commit
2e1c9b2867 ("idr: remove WARN_ON_ONCE() on negative IDs").
Then it was added back by commit 0a835c4f09 ("Reimplement IDR and IDA
using the radix tree"). However it seems that adding it back was a
mistake, given that some users such as drm_gem_handle_delete()
(DRM_IOCTL_GEM_CLOSE) pass in a value from userspace to idr_replace(),
allowing the WARN_ON_ONCE to be triggered. drm_gem_handle_delete()
actually just wants idr_replace() to return an error code if the ID is
not allocated, including in the case where the ID is invalid (negative).
So once again remove the bogus WARN_ON_ONCE().
This bug was found by syzkaller, which encountered the following
warning:
WARNING: CPU: 3 PID: 3008 at lib/idr.c:157 idr_replace+0x1d8/0x240 lib/idr.c:157
Kernel panic - not syncing: panic_on_warn set ...
CPU: 3 PID: 3008 Comm: syzkaller218828 Not tainted 4.13.0-rc4-next-20170811 #2
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
Call Trace:
fixup_bug+0x40/0x90 arch/x86/kernel/traps.c:190
do_trap_no_signal arch/x86/kernel/traps.c:224 [inline]
do_trap+0x260/0x390 arch/x86/kernel/traps.c:273
do_error_trap+0x120/0x390 arch/x86/kernel/traps.c:310
do_invalid_op+0x1b/0x20 arch/x86/kernel/traps.c:323
invalid_op+0x1e/0x30 arch/x86/entry/entry_64.S:930
RIP: 0010:idr_replace+0x1d8/0x240 lib/idr.c:157
RSP: 0018:ffff8800394bf9f8 EFLAGS: 00010297
RAX: ffff88003c6c60c0 RBX: 1ffff10007297f43 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff8800394bfa78
RBP: ffff8800394bfae0 R08: ffffffff82856487 R09: 0000000000000000
R10: ffff8800394bf9a8 R11: ffff88006c8bae28 R12: ffffffffffffffff
R13: ffff8800394bfab8 R14: dffffc0000000000 R15: ffff8800394bfbc8
drm_gem_handle_delete+0x33/0xa0 drivers/gpu/drm/drm_gem.c:297
drm_gem_close_ioctl+0xa1/0xe0 drivers/gpu/drm/drm_gem.c:671
drm_ioctl_kernel+0x1e7/0x2e0 drivers/gpu/drm/drm_ioctl.c:729
drm_ioctl+0x72e/0xa50 drivers/gpu/drm/drm_ioctl.c:825
vfs_ioctl fs/ioctl.c:45 [inline]
do_vfs_ioctl+0x1b1/0x1520 fs/ioctl.c:685
SYSC_ioctl fs/ioctl.c:700 [inline]
SyS_ioctl+0x8f/0xc0 fs/ioctl.c:691
entry_SYSCALL_64_fastpath+0x1f/0xbe
Here is a C reproducer:
#include <fcntl.h>
#include <stddef.h>
#include <stdint.h>
#include <sys/ioctl.h>
#include <drm/drm.h>
int main(void)
{
int cardfd = open("/dev/dri/card0", O_RDONLY);
ioctl(cardfd, DRM_IOCTL_GEM_CLOSE,
&(struct drm_gem_close) { .handle = -1 } );
}
Link: http://lkml.kernel.org/r/20170906235306.20534-1-ebiggers3@gmail.com
Fixes: 0a835c4f09 ("Reimplement IDR and IDA using the radix tree")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: <stable@vger.kernel.org> [v4.11+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The following new APIs are added:
int idr_alloc_ext(struct idr *idr, void *ptr, unsigned long *index,
unsigned long start, unsigned long end, gfp_t gfp);
void *idr_remove_ext(struct idr *idr, unsigned long id);
void *idr_find_ext(const struct idr *idr, unsigned long id);
void *idr_replace_ext(struct idr *idr, void *ptr, unsigned long id);
void *idr_get_next_ext(struct idr *idr, unsigned long *nextid);
Signed-off-by: Chris Mi <chrism@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Where we use the radix tree iteration macros, we need to annotate 'slot'
with __rcu. Make sure we don't forget any new places in the future with
the same CFLAGS check used for radix-tree.c.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
We can use the root entry as a bitmap and save allocating a 128 byte
bitmap for an IDA that contains only a few entries (30 on a 32-bit
machine, 62 on a 64-bit machine). This costs about 300 bytes of kernel
text on x86-64, so as long as 3 IDAs fall into this category, this
is a net win for memory consumption.
Thanks to Rasmus Villemoes for his work documenting the problem and
collecting statistics on IDAs.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
When we preload the IDA, we allocate an IDA bitmap. Instead of storing
that preallocated bitmap in the IDA, we store it in a percpu variable.
Generally there are more IDAs in the system than CPUs, so this cuts down
on the number of preallocated bitmaps that are unused, and about half
of the IDA users did not call ida_destroy() so they were leaking IDA
bitmaps.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
The IDR is very similar to the radix tree. It has some functionality that
the radix tree did not have (alloc next free, cyclic allocation, a
callback-based for_each, destroy tree), which is readily implementable on
top of the radix tree. A few small changes were needed in order to use a
tag to represent nodes with free space below them. More extensive
changes were needed to support storing NULL as a valid entry in an IDR.
Plain radix trees still interpret NULL as a not-present entry.
The IDA is reimplemented as a client of the newly enhanced radix tree. As
in the current implementation, it uses a bitmap at the last level of the
tree.
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Tested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
I wanted to wrap a bunch of ida_simple_get calls into their own locking,
until I dug around and read the original commit message. Stuff like
this should imo be added to the kernel doc, let's do that.
Link: http://lkml.kernel.org/r/20161027072216.20411-1-daniel.vetter@ffwll.ch
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__GFP_WAIT has been used to identify atomic context in callers that hold
spinlocks or are in interrupts. They are expected to be high priority and
have access one of two watermarks lower than "min" which can be referred
to as the "atomic reserve". __GFP_HIGH users get access to the first
lower watermark and can be called the "high priority reserve".
Over time, callers had a requirement to not block when fallback options
were available. Some have abused __GFP_WAIT leading to a situation where
an optimisitic allocation with a fallback option can access atomic
reserves.
This patch uses __GFP_ATOMIC to identify callers that are truely atomic,
cannot sleep and have no alternative. High priority users continue to use
__GFP_HIGH. __GFP_DIRECT_RECLAIM identifies callers that can sleep and
are willing to enter direct reclaim. __GFP_KSWAPD_RECLAIM to identify
callers that want to wake kswapd for background reclaim. __GFP_WAIT is
redefined as a caller that is willing to enter direct reclaim and wake
kswapd for background reclaim.
This patch then converts a number of sites
o __GFP_ATOMIC is used by callers that are high priority and have memory
pools for those requests. GFP_ATOMIC uses this flag.
o Callers that have a limited mempool to guarantee forward progress clear
__GFP_DIRECT_RECLAIM but keep __GFP_KSWAPD_RECLAIM. bio allocations fall
into this category where kswapd will still be woken but atomic reserves
are not used as there is a one-entry mempool to guarantee progress.
o Callers that are checking if they are non-blocking should use the
helper gfpflags_allow_blocking() where possible. This is because
checking for __GFP_WAIT as was done historically now can trigger false
positives. Some exceptions like dm-crypt.c exist where the code intent
is clearer if __GFP_DIRECT_RECLAIM is used instead of the helper due to
flag manipulations.
o Callers that built their own GFP flags instead of starting with GFP_KERNEL
and friends now also need to specify __GFP_KSWAPD_RECLAIM.
The first key hazard to watch out for is callers that removed __GFP_WAIT
and was depending on access to atomic reserves for inconspicuous reasons.
In some cases it may be appropriate for them to use __GFP_HIGH.
The second key hazard is callers that assembled their own combination of
GFP flags instead of starting with something like GFP_KERNEL. They may
now wish to specify __GFP_KSWAPD_RECLAIM. It's almost certainly harmless
if it's missed in most cases as other activity will wake kswapd.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
idr.c doesn't seem to use anything from hardirq.h (or anything included
from that). Removing it produces identical objdump -d output, and gives
44 fewer lines in the .idr.o.cmd dependency file.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch fix spelling typo found in DocBook/kernel-api.xml.
It is because the file is generated from the source comments,
I have to fix the comments in source codes.
Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
I'm working on address sanitizer project for kernel. Recently we
started experiments with stack instrumentation, to detect out-of-bounds
read/write bugs on stack.
Just after booting I've hit out-of-bounds read on stack in idr_for_each
(and in __idr_remove_all as well):
struct idr_layer **paa = &pa[0];
while (id >= 0 && id <= max) {
...
while (n < fls(id)) {
n += IDR_BITS;
p = *--paa; <--- here we are reading pa[-1] value.
}
}
Despite the fact that after this dereference we are exiting out of loop
and never use p, such behaviour is undefined and should be avoided.
Fix this by moving pointer derference to the beggining of the loop,
right before we will use it.
Signed-off-by: Andrey Ryabinin <a.ryabinin@samsung.com>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Alexey Preobrazhensky <preobr@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If "idr->hint == p" is true, it also implies "idr->hint" is true(not NULL).
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After idr subsystem is changed to RCU-awared, the free layer will not go
to the free list. The free list will not be filled up when
idr_remove(). So we don't need to shink it too.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When the smaller id is not found, idr_replace() returns -ENOENT. But
when the id is bigger enough, idr_replace() returns -EINVAL, actually
there is no difference between these two kinds of ids.
These are all unallocated id, the return values of the idr_replace() for
these ids should be the same: -ENOENT.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If the ida has at least one existing id, and when an unallocated ID
which meets a certain condition is passed to the ida_remove(), the
system will crash because it hits NULL pointer dereference.
The condition is that the unallocated ID shares the same lowest idr
layer with the existing ID, but the idr slot would be different if the
unallocated ID were to be allocated.
In this case the matching idr slot for the unallocated_id is NULL,
causing @bitmap to be NULL which the function dereferences without
checking crashing the kernel.
See the test code:
static void test3(void)
{
int id;
DEFINE_IDA(test_ida);
printk(KERN_INFO "Start test3\n");
if (ida_pre_get(&test_ida, GFP_KERNEL) < 0) return;
if (ida_get_new(&test_ida, &id) < 0) return;
ida_remove(&test_ida, 4000); /* bug: null deference here */
printk(KERN_INFO "End of test3\n");
}
It happens only when the caller tries to free an unallocated ID which is
the caller's fault. It is not a bug. But it is better to add the
proper check and complain rather than crashing the kernel.
[tj@kernel.org: updated patch description]
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If unallocated_id = (ANY * idr_max(idp->layers) + existing_id) is passed
to idr_remove(). The existing_id will be removed unexpectedly.
The following test shows this unexpected id-removal:
static void test4(void)
{
int id;
DEFINE_IDR(test_idr);
printk(KERN_INFO "Start test4\n");
id = idr_alloc(&test_idr, (void *)1, 42, 43, GFP_KERNEL);
BUG_ON(id != 42);
idr_remove(&test_idr, 42 + IDR_SIZE);
TEST_BUG_ON(idr_find(&test_idr, 42) != (void *)1);
idr_destroy(&test_idr);
printk(KERN_INFO "End of test4\n");
}
ida_remove() shares the similar problem.
It happens only when the caller tries to free an unallocated ID which is
the caller's fault. It is not a bug. But it is better to add the
proper check and complain rather than removing an existing_id silently.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
idr_replace() open-codes the logic to calculate the maximum valid ID
given the height of the idr tree; unfortunately, the open-coded logic
doesn't account for the fact that the top layer may have unused slots
and over-shifts the limit to zero when the tree is at its maximum
height.
The following test code shows it fails to replace the value for
id=((1<<27)+42):
static void test5(void)
{
int id;
DEFINE_IDR(test_idr);
#define TEST5_START ((1<<27)+42) /* use the highest layer */
printk(KERN_INFO "Start test5\n");
id = idr_alloc(&test_idr, (void *)1, TEST5_START, 0, GFP_KERNEL);
BUG_ON(id != TEST5_START);
TEST_BUG_ON(idr_replace(&test_idr, (void *)2, TEST5_START) != (void *)1);
idr_destroy(&test_idr);
printk(KERN_INFO "End of test5\n");
}
Fix the bug by using idr_max() which correctly takes into account the
maximum allowed shift.
sub_alloc() shares the same problem and may incorrectly fail with
-EAGAIN; however, this bug doesn't affect correct operation because
idr_get_empty_slot(), which already uses idr_max(), retries with the
increased @id in such cases.
[tj@kernel.org: Updated patch description.]
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Replace rcu_assign_pointer(x, NULL) with RCU_INIT_POINTER(x, NULL)
The rcu_assign_pointer() ensures that the initialization of a structure
is carried out before storing a pointer to that structure. And in the
case of the NULL pointer, there is no structure to initialize.
So, rcu_assign_pointer(p, NULL) can be safely converted to
RCU_INIT_POINTER(p, NULL)
Signed-off-by: Monam Agarwal <monamagarwal123@gmail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove no longer used deprecated code, and make local functions
static.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Jean Delvare <jdelvare@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: Philipp Reisner <philipp.reisner@linbit.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: George Spelvin <linux@horizon.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We print a dump stack after idr_remove warning. This is useful to find
the faulty piece of code. Let's do the same for ida_remove, as it would
be equally useful there.
[akpm@linux-foundation.org: convert the open-coded printk+dump_stack into WARN()]
Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As Tejun points out, there are several users of the IDR facility that
attempt to use it in a cyclic fashion. These users are likely to see
-ENOSPC errors after the counter wraps one or more times however.
This patchset adds a new idr_alloc_cyclic routine and converts several
of these users to it. Many of these users are in obscure parts of the
kernel, and I don't have a good way to test some of them. The change is
pretty straightforward though, so hopefully it won't be an issue.
There is one other cyclic user of idr_alloc that I didn't touch in
ipc/util.c. That one is doing some strange stuff that I didn't quite
understand, but it looks like it should probably be converted later
somehow.
This patch:
Thus spake Tejun Heo:
Ooh, BTW, the cyclic allocation is broken. It's prone to -ENOSPC
after the first wraparound. There are several cyclic users in the
kernel and I think it probably would be best to implement cyclic
support in idr.
This patch does that by adding new idr_alloc_cyclic function that such
users in the kernel can use. With this, there's no need for a caller to
keep track of the last value used as that's now tracked internally. This
should prevent the ENOSPC problems that can hit when the "last allocated"
counter exceeds INT_MAX.
Later patches will convert existing cyclic users to the new interface.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Eric Paris <eparis@parisplace.org>
Cc: Jack Morgenstein <jackm@dev.mellanox.co.il>
Cc: John McCutchan <john@johnmccutchan.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Robert Love <rlove@rlove.org>
Cc: Roland Dreier <roland@purestorage.com>
Cc: Sridhar Samudrala <sri@us.ibm.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Tom Tucker <tom@opengridcomputing.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
GFP_NOIO is often used for idr_alloc() inside preloaded section as the
allocation mask doesn't really matter. If the idr tree needs to be
expanded, idr_alloc() first tries to allocate using the specified
allocation mask and if it fails falls back to the preloaded buffer. This
order prevent non-preloading idr_alloc() users from taking advantage of
preloading ones by using preload buffer without filling it shifting the
burden of allocation to the preload users.
Unfortunately, this allowed/expected-to-fail kmem_cache allocation ends up
generating spurious slab lowmem warning before succeeding the request from
the preload buffer.
This patch makes idr_layer_alloc() add __GFP_NOWARN to the first
kmem_cache attempt and try kmem_cache again w/o __GFP_NOWARN after
allocation from preload_buffer fails so that lowmem warning is generated
if not suppressed by the original @gfp_mask.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: David Teigland <teigland@redhat.com>
Tested-by: David Teigland <teigland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now that all in-kernel users are converted to ues the new alloc
interface, mark the old interface deprecated. We should be able to
remove these in a few releases.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix new kernel-doc warnings in idr:
Warning(include/linux/idr.h:113): No description found for parameter 'idr'
Warning(include/linux/idr.h:113): Excess function parameter 'idp' description in 'idr_find'
Warning(lib/idr.c:232): Excess function parameter 'id' description in 'sub_alloc'
Warning(lib/idr.c:232): Excess function parameter 'id' description in 'sub_alloc'
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
idr_find(), idr_remove() and idr_replace() used to silently ignore the
sign bit and perform lookup with the rest of the bits. The weird behavior
has been changed such that negative IDs are treated as invalid. As the
behavior change was subtle, WARN_ON_ONCE() was added in the hope of
determining who's calling idr functions with negative IDs so that they can
be examined for problems.
Up until now, all two reported cases are ID number coming directly from
userland and getting fed into idr_find() and the warnings seem to cause
more problems than being helpful. Drop the WARN_ON_ONCE()s.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: <markus@trippelsdorf.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Until recently, when an negative ID is specified, idr functions used to
ignore the sign bit and proceeded with the operation with the rest of
bits, which is bizarre and error-prone. The behavior recently got changed
so that negative IDs are treated as invalid but we're triggering
WARN_ON_ONCE() on negative IDs just in case somebody was depending on the
sign bit being ignored, so that those can be detected and fixed easily.
We only need this for a while. Explain why WARN_ON_ONCE()s are there and
that they can be removed later.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
While idr lookup isn't a particularly heavy operation, it still is too
substantial to use in hot paths without worrying about the performance
implications. With recent changes, each idr_layer covers 256 slots
which should be enough to cover most use cases with single idr_layer
making lookup hint very attractive.
This patch adds idr->hint which points to the idr_layer which
allocated an ID most recently and the fast path lookup becomes
if (look up target's prefix matches that of the hinted layer)
return hint->ary[ID's offset in the leaf layer];
which can be inlined.
idr->hint is set to the leaf node on idr_fill_slot() and cleared from
free_layer().
[andriy.shevchenko@linux.intel.com: always do slow path when hint is uninitialized]
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add a field which carries the prefix of ID the idr_layer covers. This
will be used to implement lookup hint.
This patch doesn't make use of the new field and doesn't introduce any
behavior difference.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently, idr->bitmap is declared as an unsigned long which restricts
the number of bits an idr_layer can contain. All bitops can handle
arbitrary positive integer bit number and there's no reason for this
restriction.
Declare idr_layer->bitmap using DECLARE_BITMAP() instead of a single
unsigned long.
* idr_layer->bitmap is now an array. '&' dropped from params to
bitops.
* Replaced "== IDR_FULL" tests with bitmap_full() and removed
IDR_FULL.
* Replaced find_next_bit() on ~bitmap with find_next_zero_bit().
* Replaced "bitmap = 0" with bitmap_clear().
This patch doesn't (or at least shouldn't) introduce any behavior
changes.
[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
MAX_IDR_MASK is another weirdness in the idr interface. As idr covers
whole positive integer range, it's defined as 0x7fffffff or INT_MAX.
Its usage in idr_find(), idr_replace() and idr_remove() is bizarre.
They basically mask off the sign bit and operate on the rest, so if
the caller, by accident, passes in a negative number, the sign bit
will be masked off and the remaining part will be used as if that was
the input, which is worse than crashing.
The constant is visible in idr.h and there are several users in the
kernel.
* drivers/i2c/i2c-core.c:i2c_add_numbered_adapter()
Basically used to test if adap->nr is a negative number which isn't
-1 and returns -EINVAL if so. idr_alloc() already has negative
@start checking (w/ WARN_ON_ONCE), so this can go away.
* drivers/infiniband/core/cm.c:cm_alloc_id()
drivers/infiniband/hw/mlx4/cm.c:id_map_alloc()
Used to wrap cyclic @start. Can be replaced with max(next, 0).
Note that this type of cyclic allocation using idr is buggy. These
are prone to spurious -ENOSPC failure after the first wraparound.
* fs/super.c:get_anon_bdev()
The ID allocated from ida is masked off before being tested whether
it's inside valid range. ida allocated ID can never be a negative
number and the masking is unnecessary.
Update idr_*() functions to fail with -EINVAL when negative @id is
specified and update other MAX_IDR_MASK users as described above.
This leaves MAX_IDR_MASK without any user, remove it and relocate
other MAX_IDR_* constants to lib/idr.c.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jean Delvare <khali@linux-fr.org>
Cc: Roland Dreier <roland@kernel.org>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Hal Rosenstock <hal.rosenstock@gmail.com>
Cc: "Marciniszyn, Mike" <mike.marciniszyn@intel.com>
Cc: Jack Morgenstein <jackm@dev.mellanox.co.il>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Wolfram Sang <wolfram@the-dreams.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Most functions in idr fail to deal with the high bits when the idr
tree grows to the maximum height.
* idr_get_empty_slot() stops growing idr tree once the depth reaches
MAX_IDR_LEVEL - 1, which is one depth shallower than necessary to
cover the whole range. The function doesn't even notice that it
didn't grow the tree enough and ends up allocating the wrong ID
given sufficiently high @starting_id.
For example, on 64 bit, if the starting id is 0x7fffff01,
idr_get_empty_slot() will grow the tree 5 layer deep, which only
covers the 30 bits and then proceed to allocate as if the bit 30
wasn't specified. It ends up allocating 0x3fffff01 without the bit
30 but still returns 0x7fffff01.
* __idr_remove_all() will not remove anything if the tree is fully
grown.
* idr_find() can't find anything if the tree is fully grown.
* idr_for_each() and idr_get_next() can't iterate anything if the tree
is fully grown.
Fix it by introducing idr_max() which returns the maximum possible ID
given the depth of tree and replacing the id limit checks in all
affected places.
As the idr_layer pointer array pa[] needs to be 1 larger than the
maximum depth, enlarge pa[] arrays by one.
While this plugs the discovered issues, the whole code base is
horrible and in desparate need of rewrite. It's fragile like hell,
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current idr interface is very cumbersome.
* For all allocations, two function calls - idr_pre_get() and
idr_get_new*() - should be made.
* idr_pre_get() doesn't guarantee that the following idr_get_new*()
will not fail from memory shortage. If idr_get_new*() returns
-EAGAIN, the caller is expected to retry pre_get and allocation.
* idr_get_new*() can't enforce upper limit. Upper limit can only be
enforced by allocating and then freeing if above limit.
* idr_layer buffer is unnecessarily per-idr. Each idr ends up keeping
around MAX_IDR_FREE idr_layers. The memory consumed per idr is
under two pages but it makes it difficult to make idr_layer larger.
This patch implements the following new set of allocation functions.
* idr_preload[_end]() - Similar to radix preload but doesn't fail.
The first idr_alloc() inside preload section can be treated as if it
were called with @gfp_mask used for idr_preload().
* idr_alloc() - Allocate an ID w/ lower and upper limits. Takes
@gfp_flags and can be used w/o preloading. When used inside
preloaded section, the allocation mask of preloading can be assumed.
If idr_alloc() can be called from a context which allows sufficiently
relaxed @gfp_mask, it can be used by itself. If, for example,
idr_alloc() is called inside spinlock protected region, preloading can
be used like the following.
idr_preload(GFP_KERNEL);
spin_lock(lock);
id = idr_alloc(idr, ptr, start, end, GFP_NOWAIT);
spin_unlock(lock);
idr_preload_end();
if (id < 0)
error;
which is much simpler and less error-prone than idr_pre_get and
idr_get_new*() loop.
The new interface uses per-pcu idr_layer buffer and thus the number of
idr's in the system doesn't affect the amount of memory used for
preloading.
idr_layer_alloc() is introduced to handle idr_layer allocations for
both old and new ID allocation paths. This is a bit hairy now but the
new interface is expected to replace the old and the internal
implementation eventually will become simpler.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move slot filling to idr_fill_slot() from idr_get_new_above_int() and
make idr_get_new_above() directly call it. idr_get_new_above_int() is
no longer needed and removed.
This will be used to implement a new ID allocation interface.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
idr uses -1, IDR_NEED_TO_GROW and IDR_NOMORE_SPACE to communicate
exception conditions internally. The return value is later translated
to errno values using _idr_rc_to_errno().
This is confusing. Drop the custom ones and consistently use -EAGAIN
for "tree needs to grow", -ENOMEM for "need more memory" and -ENOSPC for
"ran out of ID space".
Due to the weird memory preloading mechanism, [ra]_get_new*() return
-EAGAIN on memory shortage, so we need to substitute -ENOMEM w/
-EAGAIN on those interface functions. They'll eventually be cleaned
up and the translations will go away.
This patch doesn't introduce any functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* Move idr_for_each_entry() definition next to other idr related
definitions.
* Make id[r|a]_get_new() inline wrappers of id[r|a]_get_new_above().
This changes the implementation of idr_get_new() but the new
implementation is trivial. This patch doesn't introduce any
functional change.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There was only one legitimate use of idr_remove_all() and a lot more of
incorrect uses (or lack of it). Now that idr_destroy() implies
idr_remove_all() and all the in-kernel users updated not to use it,
there's no reason to keep it around. Mark it deprecated so that we can
later unexport it.
idr_remove_all() is made an inline function calling __idr_remove_all()
to avoid triggering deprecated warning on EXPORT_SYMBOL().
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
idr is silly in quite a few ways, one of which is how it's supposed to
be destroyed - idr_destroy() doesn't release IDs and doesn't even whine
if the idr isn't empty. If the caller forgets idr_remove_all(), it
simply leaks memory.
Even ida gets this wrong and leaks memory on destruction. There is
absoltely no reason not to call idr_remove_all() from idr_destroy().
Nobody is abusing idr_destroy() for shrinking free layer buffer and
continues to use idr after idr_destroy(), so it's safe to do remove_all
from destroy.
In the whole kernel, there is only one place where idr_remove_all() is
legitimiately used without following idr_destroy() while there are quite
a few places where the caller forgets either idr_remove_all() or
idr_destroy() leaking memory.
This patch makes idr_destroy() call idr_destroy_all() and updates the
function description accordingly.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Matthew Wilcox
Rasmus Villemoes
Mel Gorman
Eric Biggers
Chris Mi
Daniel Vetter
Linus Torvalds
Masanari Iida
Andrey Ryabinin
Lai Jiangshan
Monam Agarwal
Stephen Hemminger
Andreas Gruenbacher
Jean Delvare
Jeff Layton
Tejun Heo
Randy Dunlap