2014-12-08 18:35:51 +00:00
|
|
|
/*
|
2010-05-17 22:18:00 +00:00
|
|
|
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
|
|
|
|
|
* Copyright (C) 2007 The Regents of the University of California.
|
|
|
|
|
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
|
|
|
|
|
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
|
2008-05-26 04:38:26 +00:00
|
|
|
* UCRL-CODE-235197
|
|
|
|
|
*
|
2010-05-17 22:18:00 +00:00
|
|
|
* This file is part of the SPL, Solaris Porting Layer.
|
2013-03-05 01:26:55 +00:00
|
|
|
* For details, see <http://zfsonlinux.org/>.
|
2010-05-17 22:18:00 +00:00
|
|
|
*
|
|
|
|
|
* The SPL is free software; you can redistribute it and/or modify it
|
|
|
|
|
* under the terms of the GNU General Public License as published by the
|
|
|
|
|
* Free Software Foundation; either version 2 of the License, or (at your
|
|
|
|
|
* option) any later version.
|
2008-05-26 04:38:26 +00:00
|
|
|
*
|
2010-05-17 22:18:00 +00:00
|
|
|
* The SPL is distributed in the hope that it will be useful, but WITHOUT
|
2008-05-26 04:38:26 +00:00
|
|
|
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
|
|
|
* for more details.
|
|
|
|
|
*
|
|
|
|
|
* You should have received a copy of the GNU General Public License along
|
2010-05-17 22:18:00 +00:00
|
|
|
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
|
2014-12-08 18:35:51 +00:00
|
|
|
*/
|
2008-05-26 04:38:26 +00:00
|
|
|
|
2008-02-28 00:52:31 +00:00
|
|
|
#ifndef _SPL_KMEM_H
|
|
|
|
|
#define _SPL_KMEM_H
|
2008-02-26 20:36:04 +00:00
|
|
|
|
2014-07-13 18:45:20 +00:00
|
|
|
#include <sys/debug.h>
|
2008-02-26 20:36:04 +00:00
|
|
|
#include <linux/slab.h>
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
#include <linux/sched.h>
|
2019-07-21 17:34:07 +00:00
|
|
|
#include <linux/mm.h>
|
|
|
|
|
#include <linux/vmalloc.h>
|
2014-12-08 18:04:42 +00:00
|
|
|
|
|
|
|
|
extern int kmem_debugging(void);
|
|
|
|
|
extern char *kmem_vasprintf(const char *fmt, va_list ap);
|
|
|
|
|
extern char *kmem_asprintf(const char *fmt, ...);
|
2019-10-10 16:47:06 +00:00
|
|
|
extern char *kmem_strdup(const char *str);
|
|
|
|
|
extern void kmem_strfree(char *str);
|
2008-11-03 20:34:17 +00:00
|
|
|
|
2008-02-26 20:36:04 +00:00
|
|
|
/*
|
|
|
|
|
* Memory allocation interfaces
|
|
|
|
|
*/
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
#define KM_SLEEP 0x0000 /* can block for memory; success guaranteed */
|
|
|
|
|
#define KM_NOSLEEP 0x0001 /* cannot block for memory; may fail */
|
|
|
|
|
#define KM_PUSHPAGE 0x0004 /* can block for memory; may use reserve */
|
|
|
|
|
#define KM_ZERO 0x1000 /* zero the allocation */
|
|
|
|
|
#define KM_VMEM 0x2000 /* caller is vmem_* wrapper */
|
2008-02-26 20:36:04 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
#define KM_PUBLIC_MASK (KM_SLEEP | KM_NOSLEEP | KM_PUSHPAGE)
|
2008-08-11 22:13:47 +00:00
|
|
|
|
2017-05-09 17:36:54 +00:00
|
|
|
static int spl_fstrans_check(void);
|
2019-07-21 17:34:07 +00:00
|
|
|
void *spl_kvmalloc(size_t size, gfp_t flags);
|
2017-05-09 17:36:54 +00:00
|
|
|
|
2009-11-12 23:11:24 +00:00
|
|
|
/*
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
* Convert a KM_* flags mask to its Linux GFP_* counterpart. The conversion
|
|
|
|
|
* function is context aware which means that KM_SLEEP allocations can be
|
|
|
|
|
* safely used in syncing contexts which have set PF_FSTRANS.
|
2009-11-12 23:11:24 +00:00
|
|
|
*/
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
static inline gfp_t
|
|
|
|
|
kmem_flags_convert(int flags)
|
2009-11-12 23:11:24 +00:00
|
|
|
{
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
gfp_t lflags = __GFP_NOWARN | __GFP_COMP;
|
2009-11-12 23:11:24 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
if (flags & KM_NOSLEEP) {
|
|
|
|
|
lflags |= GFP_ATOMIC | __GFP_NORETRY;
|
|
|
|
|
} else {
|
|
|
|
|
lflags |= GFP_KERNEL;
|
2017-05-09 17:36:54 +00:00
|
|
|
if (spl_fstrans_check())
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
lflags &= ~(__GFP_IO|__GFP_FS);
|
|
|
|
|
}
|
2009-11-12 23:11:24 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
if (flags & KM_PUSHPAGE)
|
|
|
|
|
lflags |= __GFP_HIGH;
|
2009-11-12 23:11:24 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
if (flags & KM_ZERO)
|
|
|
|
|
lflags |= __GFP_ZERO;
|
2009-11-12 23:11:24 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
return (lflags);
|
2010-07-26 22:47:55 +00:00
|
|
|
}
|
|
|
|
|
|
2014-07-13 18:45:20 +00:00
|
|
|
typedef struct {
|
|
|
|
|
struct task_struct *fstrans_thread;
|
|
|
|
|
unsigned int saved_flags;
|
|
|
|
|
} fstrans_cookie_t;
|
|
|
|
|
|
2017-05-09 17:36:54 +00:00
|
|
|
/*
|
|
|
|
|
* Introduced in Linux 3.9, however this cannot be solely relied on before
|
|
|
|
|
* Linux 3.18 as it doesn't turn off __GFP_FS as it should.
|
|
|
|
|
*/
|
2016-01-18 22:41:45 +00:00
|
|
|
#ifdef PF_MEMALLOC_NOIO
|
2017-05-09 17:36:54 +00:00
|
|
|
#define __SPL_PF_MEMALLOC_NOIO (PF_MEMALLOC_NOIO)
|
|
|
|
|
#else
|
|
|
|
|
#define __SPL_PF_MEMALLOC_NOIO (0)
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* PF_FSTRANS is removed from Linux 4.12
|
|
|
|
|
*/
|
|
|
|
|
#ifdef PF_FSTRANS
|
|
|
|
|
#define __SPL_PF_FSTRANS (PF_FSTRANS)
|
2016-01-18 22:41:45 +00:00
|
|
|
#else
|
2017-05-09 17:36:54 +00:00
|
|
|
#define __SPL_PF_FSTRANS (0)
|
2016-01-18 22:41:45 +00:00
|
|
|
#endif
|
|
|
|
|
|
2017-05-09 17:36:54 +00:00
|
|
|
#define SPL_FSTRANS (__SPL_PF_FSTRANS|__SPL_PF_MEMALLOC_NOIO)
|
|
|
|
|
|
2014-07-13 18:45:20 +00:00
|
|
|
static inline fstrans_cookie_t
|
|
|
|
|
spl_fstrans_mark(void)
|
|
|
|
|
{
|
|
|
|
|
fstrans_cookie_t cookie;
|
|
|
|
|
|
2017-05-09 17:36:54 +00:00
|
|
|
BUILD_BUG_ON(SPL_FSTRANS == 0);
|
|
|
|
|
|
2014-07-13 18:45:20 +00:00
|
|
|
cookie.fstrans_thread = current;
|
2016-01-18 22:41:45 +00:00
|
|
|
cookie.saved_flags = current->flags & SPL_FSTRANS;
|
|
|
|
|
current->flags |= SPL_FSTRANS;
|
2014-07-13 18:45:20 +00:00
|
|
|
|
|
|
|
|
return (cookie);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
|
spl_fstrans_unmark(fstrans_cookie_t cookie)
|
|
|
|
|
{
|
|
|
|
|
ASSERT3P(cookie.fstrans_thread, ==, current);
|
2016-01-18 22:41:45 +00:00
|
|
|
ASSERT((current->flags & SPL_FSTRANS) == SPL_FSTRANS);
|
2014-07-13 18:45:20 +00:00
|
|
|
|
2016-01-18 22:41:45 +00:00
|
|
|
current->flags &= ~SPL_FSTRANS;
|
2014-07-13 18:45:20 +00:00
|
|
|
current->flags |= cookie.saved_flags;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline int
|
|
|
|
|
spl_fstrans_check(void)
|
|
|
|
|
{
|
2017-05-09 17:36:54 +00:00
|
|
|
return (current->flags & SPL_FSTRANS);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* specifically used to check PF_FSTRANS flag, cannot be relied on for
|
|
|
|
|
* checking spl_fstrans_mark().
|
|
|
|
|
*/
|
|
|
|
|
static inline int
|
|
|
|
|
__spl_pf_fstrans_check(void)
|
|
|
|
|
{
|
|
|
|
|
return (current->flags & __SPL_PF_FSTRANS);
|
2014-07-13 18:45:20 +00:00
|
|
|
}
|
|
|
|
|
|
2014-12-08 18:35:51 +00:00
|
|
|
#ifdef HAVE_ATOMIC64_T
|
|
|
|
|
#define kmem_alloc_used_add(size) atomic64_add(size, &kmem_alloc_used)
|
|
|
|
|
#define kmem_alloc_used_sub(size) atomic64_sub(size, &kmem_alloc_used)
|
|
|
|
|
#define kmem_alloc_used_read() atomic64_read(&kmem_alloc_used)
|
|
|
|
|
#define kmem_alloc_used_set(size) atomic64_set(&kmem_alloc_used, size)
|
2010-07-26 22:47:55 +00:00
|
|
|
extern atomic64_t kmem_alloc_used;
|
2009-12-04 23:54:12 +00:00
|
|
|
extern unsigned long long kmem_alloc_max;
|
2014-12-08 18:35:51 +00:00
|
|
|
#else /* HAVE_ATOMIC64_T */
|
|
|
|
|
#define kmem_alloc_used_add(size) atomic_add(size, &kmem_alloc_used)
|
|
|
|
|
#define kmem_alloc_used_sub(size) atomic_sub(size, &kmem_alloc_used)
|
|
|
|
|
#define kmem_alloc_used_read() atomic_read(&kmem_alloc_used)
|
|
|
|
|
#define kmem_alloc_used_set(size) atomic_set(&kmem_alloc_used, size)
|
2010-07-26 22:47:55 +00:00
|
|
|
extern atomic_t kmem_alloc_used;
|
|
|
|
|
extern unsigned long long kmem_alloc_max;
|
2014-12-08 18:35:51 +00:00
|
|
|
#endif /* HAVE_ATOMIC64_T */
|
2008-11-03 21:06:04 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
extern unsigned int spl_kmem_alloc_warn;
|
|
|
|
|
extern unsigned int spl_kmem_alloc_max;
|
2008-02-26 20:36:04 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
#define kmem_alloc(sz, fl) spl_kmem_alloc((sz), (fl), __func__, __LINE__)
|
|
|
|
|
#define kmem_zalloc(sz, fl) spl_kmem_zalloc((sz), (fl), __func__, __LINE__)
|
|
|
|
|
#define kmem_free(ptr, sz) spl_kmem_free((ptr), (sz))
|
2017-03-15 23:41:52 +00:00
|
|
|
#define kmem_cache_reap_active spl_kmem_cache_reap_active
|
2008-02-26 20:36:04 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
extern void *spl_kmem_alloc(size_t sz, int fl, const char *func, int line);
|
|
|
|
|
extern void *spl_kmem_zalloc(size_t sz, int fl, const char *func, int line);
|
|
|
|
|
extern void spl_kmem_free(const void *ptr, size_t sz);
|
2008-03-18 04:56:43 +00:00
|
|
|
|
Refactor generic memory allocation interfaces
This patch achieves the following goals:
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2014-12-08 20:37:14 +00:00
|
|
|
/*
|
|
|
|
|
* The following functions are only available for internal use.
|
|
|
|
|
*/
|
|
|
|
|
extern void *spl_kmem_alloc_impl(size_t size, int flags, int node);
|
|
|
|
|
extern void *spl_kmem_alloc_debug(size_t size, int flags, int node);
|
|
|
|
|
extern void *spl_kmem_alloc_track(size_t size, int flags,
|
|
|
|
|
const char *func, int line, int node);
|
|
|
|
|
extern void spl_kmem_free_impl(const void *buf, size_t size);
|
|
|
|
|
extern void spl_kmem_free_debug(const void *buf, size_t size);
|
|
|
|
|
extern void spl_kmem_free_track(const void *buf, size_t size);
|
|
|
|
|
|
|
|
|
|
extern int spl_kmem_init(void);
|
|
|
|
|
extern void spl_kmem_fini(void);
|
2017-03-15 23:41:52 +00:00
|
|
|
extern int spl_kmem_cache_reap_active(void);
|
2008-02-26 20:36:04 +00:00
|
|
|
|
2008-02-28 00:52:31 +00:00
|
|
|
#endif /* _SPL_KMEM_H */
|
