Mark those kmem allocations that are known to be easily triggered from
userspace as __GFP_ACCOUNT/SLAB_ACCOUNT, which makes them accounted to
memcg. For the list, see below:
- threadinfo
- task_struct
- task_delay_info
- pid
- cred
- mm_struct
- vm_area_struct and vm_region (nommu)
- anon_vma and anon_vma_chain
- signal_struct
- sighand_struct
- fs_struct
- files_struct
- fdtable and fdtable->full_fds_bits
- dentry and external_name
- inode for all filesystems. This is the most tedious part, because
most filesystems overwrite the alloc_inode method.
The list is far from complete, so feel free to add more objects.
Nevertheless, it should be close to "account everything" approach and
keep most workloads within bounds. Malevolent users will be able to
breach the limit, but this was possible even with the former "account
everything" approach (simply because it did not account everything in
fact).
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
that's the bulk of filesystem drivers dealing with inodes of their own
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Since "BDI: Provide backing device capability information [try #3]" the
backing_dev_info structure also provides flags for the kind of mmap
operation available in a nommu environment, which is entirely unrelated
to it's original purpose.
Introduce a new nommu-only file operation to provide this information to
the nommu mmap code instead. Splitting this from the backing_dev_info
structure allows to remove lots of backing_dev_info instance that aren't
otherwise needed, and entirely gets rid of the concept of providing a
backing_dev_info for a character device. It also removes the need for
the mtd_inodefs filesystem.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Tejun Heo <tj@kernel.org>
Acked-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Mark function as static in exofs/super.c because it is not used outside
this file.
This also eliminates the following warning in exofs/super.c:
fs/exofs/super.c:546:5: warning: no previous prototype \
for __alloc_dev_table[-Wmissing-prototypes]
Signed-off-by: Rashika Kheria <rashika.kheria@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Modify the request_module to prefix the file system type with "fs-"
and add aliases to all of the filesystems that can be built as modules
to match.
A common practice is to build all of the kernel code and leave code
that is not commonly needed as modules, with the result that many
users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible
modules that can be loaded by mount to just filesystems trivially
making things safer with no real cost.
Using aliases means user space can control the policy of which
filesystem modules are auto-loaded by editing /etc/modprobe.d/*.conf
with blacklist and alias directives. Allowing simple, safe,
well understood work-arounds to known problematic software.
This also addresses a rare but unfortunate problem where the filesystem
name is not the same as it's module name and module auto-loading
would not work. While writing this patch I saw a handful of such
cases. The most significant being autofs that lives in the module
autofs4.
This is relevant to user namespaces because we can reach the request
module in get_fs_type() without having any special permissions, and
people get uncomfortable when a user specified string (in this case
the filesystem type) goes all of the way to request_module.
After having looked at this issue I don't think there is any
particular reason to perform any filtering or permission checks beyond
making it clear in the module request that we want a filesystem
module. The common pattern in the kernel is to call request_module()
without regards to the users permissions. In general all a filesystem
module does once loaded is call register_filesystem() and go to sleep.
Which means there is not much attack surface exposed by loading a
filesytem module unless the filesystem is mounted. In a user
namespace filesystems are not mounted unless .fs_flags = FS_USERNS_MOUNT,
which most filesystems do not set today.
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reported-by: Kees Cook <keescook@google.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
There's no reason to call rcu_barrier() on every
deactivate_locked_super(). We only need to make sure that all delayed rcu
free inodes are flushed before we destroy related cache.
Removing rcu_barrier() from deactivate_locked_super() affects some fast
paths. E.g. on my machine exit_group() of a last process in IPC
namespace takes 0.07538s. rcu_barrier() takes 0.05188s of that time.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Exofs has the '->write_super()' handler and makes some use of the '->s_dirt'
superblock flag, but it really needs neither of them because it never sets
's_dirt' to one which means the VFS never calls its '->write_super()' handler.
Thus, remove both.
Note, I am trying to remove both 's_dirt' and 'write_super()' from VFS
altogether once all users are gone.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Introduce sysfs infrastructure for exofs cluster filesystem.
Each OSD target shows up as below in the sysfs hierarchy:
/sys/fs/exofs/<osdname>_<partition_id>/devX
Where <osdname>_<partition_id> is the unique identification
of a Superblock.
Where devX: 0 <= X < device_table_size. They are ordered
in device-table order as specified to the mkfs.exofs command
Each OSD device devX has following attributes :
osdname - ReadOnly
systemid - ReadOnly
uri - Read/Write
It is up to user-mode to update devX/uri for support of
autologin.
These sysfs information are used both for autologin as well
as support for exporting exofs via a pNFSD server in user-mode.
(.eg NFS-Ganesha)
Signed-off-by: Sachin Bhamare <sbhamare@panasas.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
If at exofs_fill_super() we had an early termination
do to any error, like an IO error while reading the
super-block. We would crash inside exofs_free_sbi().
This is because sbi->oc.numdevs was set to 1, before
we actually have a device table at all.
Fix it by moving the sbi->oc.numdevs = 1 to after the
allocation of the device table.
Reported-by: Johannes Schild <JSchild@gmx.de>
Stable: This is a bug since v3.2.0
CC: Stable Tree <stable@kernel.org>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Pull trivial exofs changes from Boaz Harrosh:
"Just nothingness really. The big exofs changes are reserved for the
next merge window."
* 'for-linus' of git://git.open-osd.org/linux-open-osd:
exofs: Cap on the memcpy() size
exofs: (trivial) Fix typo in super.c
exofs: fix endian conversion in exofs_sync_fs()
New field of struct super_block - ->s_max_links. Maximal allowed
value of ->i_nlink or 0; in the latter case all checks still need
to be done in ->link/->mkdir/->rename instances. Note that this
limit applies both to directoris and to non-directories.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This data comes from the device, so probably it's fairly trustworthy but
it makes the static checkers happy if we check it.
[Boaz]
the system_id_len is zero, if not present, or always OSD_SYSTEMID_LEN.
So always copy OSD_SYSTEMID_LEN bytes.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
fscb->s_numfiles is an __le64 field so we need to use cpu_to_le64()
to get a little endian 64 bit on big endian systems.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
We have already set ->s_root, so ->put_super() is going to be called.
Freeing ->s_fs_info is a bloody bad idea when it's going to be
dereferenced very shortly...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Seeing that just about every destructor got that INIT_LIST_HEAD() copied into
it, there is no point whatsoever keeping this INIT_LIST_HEAD in inode_init_once();
the cost of taking it into inode_init_always() will be negligible for pipes
and sockets and negative for everything else. Not to mention the removal of
boilerplate code from ->destroy_inode() instances...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Some files were using the complete module.h infrastructure without
actually including the header at all. Fix them up in advance so
once the implicit presence is removed, we won't get failures like this:
CC [M] fs/nfsd/nfssvc.o
fs/nfsd/nfssvc.c: In function 'nfsd_create_serv':
fs/nfsd/nfssvc.c:335: error: 'THIS_MODULE' undeclared (first use in this function)
fs/nfsd/nfssvc.c:335: error: (Each undeclared identifier is reported only once
fs/nfsd/nfssvc.c:335: error: for each function it appears in.)
fs/nfsd/nfssvc.c: In function 'nfsd':
fs/nfsd/nfssvc.c:555: error: implicit declaration of function 'module_put_and_exit'
make[3]: *** [fs/nfsd/nfssvc.o] Error 1
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
All users of the ore will need to check if current code
supports the given layout. For example RAID5/6 is not
currently supported.
So move all the checks from exofs/super.c to a new
ore_verify_layout() to be used by ore users.
Note that any new layout should be passed through the
ore_verify_layout() because the ore engine will prepare
and verify some internal members of ore_layout, and
assumes it's called.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
In the pNFS obj-LD the device table at the layout level needs
to point to a device_cache node, where it is possible and likely
that many layouts will point to the same device-nodes.
In Exofs we have a more orderly structure where we have a single
array of devices that repeats twice for a round-robin view of the
device table
This patch moves to a model that can be used by the pNFS obj-LD
where struct ore_components holds an array of ore_dev-pointers.
(ore_dev is newly defined and contains a struct osd_dev *od
member)
Each pointer in the array of pointers will point to a bigger
user-defined dev_struct. That can be accessed by use of the
container_of macro.
In Exofs an __alloc_dev_table() function allocates the
ore_dev-pointers array as well as an exofs_dev array, in one
allocation and does the addresses dance to set everything pointing
correctly. It still keeps the double allocation trick for the
inodes round-robin view of the table.
The device table is always allocated dynamically, also for the
single device case. So it is unconditionally freed at umount.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
The struct pnfs_osd_data_map data_map member of exofs_sb_info was
never used after mount. In fact all it's members were duplicated
by the ore_layout structure. So just remove the duplicated information.
Also removed some stupid, but perfectly supported, restrictions on
layout parameters. The case where num_devices is not divisible by
mirror_count+1 is perfectly fine since the rotating device view
will eventually use all the devices it can get.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: Benny Halevy <bhalevy@tonian.com>
ore_components already has a comps member so this leads
to things like comps->comps which is annoying. the name oc
was already used in new code. So rename all old usage of
ore_components comps => ore_components oc.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
This quiets the following sparse noise:
warning: symbol 'exofs_sync_fs' was not declared. Should it be static?
warning: symbol 'exofs_free_sbi' was not declared. Should it be static?
warning: symbol 'exofs_get_parent' was not declared. Should it be static?
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
ORE stands for "Objects Raid Engine"
This patch is a mechanical rename of everything that was in ios.c
and its API declaration to an ore.c and an osd_ore.h header. The ore
engine will later be used by the pnfs objects layout driver.
* File ios.c => ore.c
* Declaration of types and API are moved from exofs.h to a new
osd_ore.h
* All used types are prefixed by ore_ from their exofs_ name.
* Shift includes from exofs.h to osd_ore.h so osd_ore.h is
independent, include it from exofs.h.
Other than a pure rename there are no other changes. Next patch
will move the ore into it's own module and will export the API
to be used by exofs and later the layout driver
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Exofs raid engine was saving on memory space by having a single layout-info,
single pid, and a single device-table, global to the filesystem. Then passing
a credential and object_id info at the io_state level, private for each
inode. It would also devise this contraption of rotating the device table
view for each inode->ino to spread out the device usage.
This is not compatible with the pnfs-objects standard, demanding that
each inode can have it's own layout-info, device-table, and each object
component it's own pid, oid and creds.
So: Bring exofs raid engine to be usable for generic pnfs-objects use by:
* Define an exofs_comp structure that holds obj_id and credential info.
* Break up exofs_layout struct to an exofs_components structure that holds a
possible array of exofs_comp and the array of devices + the size of the
arrays.
* Add a "comps" parameter to get_io_state() that specifies the ids creds
and device array to use for each IO.
This enables to keep the layout global, but the device-table view, creds
and IDs at the inode level. It only adds two 64bit to each inode, since
some of these members already existed in another form.
* ios raid engine now access layout-info and comps-info through the passed
pointers. Everything is pre-prepared by caller for generic access of
these structures and arrays.
At the exofs Level:
* Super block holds an exofs_components struct that holds the device
array, previously in layout. The devices there are in device-table
order. The device-array is twice bigger and repeats the device-table
twice so now each inode's device array can point to a random device
and have a round-robin view of the table, making it compatible to
previous exofs versions.
* Each inode has an exofs_components struct that is initialized at
load time, with it's own view of the device table IDs and creds.
When doing IO this gets passed to the io_state together with the
layout.
While preforming this change. Bugs where found where credentials with the
wrong IDs where used to access the different SB objects (super.c). As well
as some dead code. It was never noticed because the target we use does not
check the credentials.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
ios.c will be moving to an external library, for use by the
objects-layout-driver. Remove from it some exofs specific functions.
Also g_attr_logical_length is used both by inode.c and ios.c
move definition to the later, to keep it independent
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Since the beginning we realloced the sbi structure when a bigger
then one device table was specified. (I know that was really stupid).
Then much later when "register bdi" was added (By Jens) it was
registering the pointer to sbi->bdi before the realloc.
We never saw this problem because up till now the realloc did not
do anything since the device table was small enough to fit in the
original allocation. But once we starting testing with large device
tables (Bigger then 28) we noticed the crash of writeback operating
on a deallocated pointer.
* Avoid the all mess by allocating the device-table as a second array
and get rid of the variable-sized structure and the rest of this
mess.
* Take the chance to clean near by structures and comments.
* Add a needed dprint on startup to indicate the loaded layout.
* Also move the bdi registration to the very end because it will
only fail in a low memory, which will probably fail before hand.
There are many more likely causes to not load before that. This
way the error handling is made simpler. (Just doing this would be
enough to fix the BUG)
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
One leftover from the days of IBM's original code, is an SB counter
that counts in-flight asynchronous commands. And a piece of code that
waits for the counter to reach zero at unmount. I guess it might have
been needed then, cause of some reference missing or something.
I'm not removing it yet but am putting a warning message if ever this
counter triggers at unmount. If I'll never see it triggers or reported
I'll remove the counter for good.
(I had this print as a debug output for a long time and never had it
trigger)
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Before when creating a new inode, we'd set the sb->s_dirt flag,
and sometime later the system would write out s_nextid as part
of the sb_info. Also on inode sync we would force the sb sync
as well.
Define the s_nextid as a new partition attribute and set it
every time we create a new object.
At mount we read it from it's new place.
We now never set sb->s_dirt anywhere in exofs. write_super
is actually never called. The call to exofs_write_super from
exofs_put_super is also removed because the VFS always calls
->sync_fs before calling ->put_super twice.
To stay backward-and-forward compatible we also write the old
s_nextid in the super_block object at unmount, and support zero
length attribute on mount.
This also fixes a BUG where in layouts when group_width was not
a divisor of EXOFS_SUPER_ID (0x10000) the s_nextid was not read
from the device it was written to. Because of the sliding window
layout trick, and because the read was always done from the 0
device but the write was done via the raid engine that might slide
the device view. Now we read and write through the raid engine.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
If /dev/osd* devices are shuffled because more devices
where added, and/or login order has changed. It is hard to
mount the FS you want.
Add an option to mount by osdname. osdname is any osd-device's
osdname as specified to the mkfs.exofs command when formatting
the osd-devices.
The new mount format is:
OPT="osdname=$UUID0,pid=$PID,_netdev"
mount -t exofs -o $OPT $DEV_OSD0 $MOUNTDIR
if "osdname=" is specified in options above $DEV_OSD0 is
ignored and can be empty.
Also while at it: Removed some old unused Opt_* enums.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
* Set all inode->i_mapping->backing_dev_info to point to
the per super-block sb->s_bdi.
* Calculating a read_ahead that is:
- preferable 2 stripes long
(Future patch will add a mount option to override this)
- Minimum 128K aligned up to stripe-size
- Caped to maximum-IO-sizes round down to stripe_size.
(Max sizes are governed by max bio-size that fits in a page
times number-of-devices)
CC: Marc Dionne <marc.c.dionne@gmail.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
RCU free the struct inode. This will allow:
- Subsequent store-free path walking patch. The inode must be consulted for
permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
to take i_lock no longer need to take sb_inode_list_lock to walk the list in
the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
page lock to follow page->mapping.
The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.
In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.
The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
* 'for-linus' of git://git.open-osd.org/linux-open-osd:
exofs: Fix groups code when num_devices is not divisible by group_width
exofs: Remove useless optimization
exofs: exofs_file_fsync and exofs_file_flush correctness
exofs: Remove superfluous dependency on buffer_head and writeback
exofs_releasepage && exofs_invalidatepage are never called.
Leave the WARN_ONs but remove any code. Remove the
cleanup other stale #includes.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
* _calc_stripe_info() changes to accommodate for grouping
calculations. Returns additional information
* old _prepare_pages() becomes _prepare_one_group()
which stores pages belonging to one device group.
* New _prepare_for_striping iterates on all groups calling
_prepare_one_group().
* Enable mounting of groups data_maps (group_width != 0)
[QUESTION]
what is faster A or B;
A. x += stride;
x = x % width + first_x;
B x += stride
if (x < last_x)
x = first_x;
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
We now support striping over mirror devices. Including variable sized
stripe_unit.
Some limits:
* stripe_unit must be a multiple of PAGE_SIZE
* stripe_unit * stripe_count is maximum upto 32-bit (4Gb)
Tested RAID0 over mirrors, RAID0 only, mirrors only. All check.
Design notes:
* I'm not using a vectored raid-engine mechanism yet. Following the
pnfs-objects-layout data-map structure, "Mirror" is just a private
case of "group_width" == 1, and RAID0 is a private case of
"Mirrors" == 1. The performance lose of the general case over the
particular special case optimization is totally negligible, also
considering the extra code size.
* In general I added a prepare_stripes() stage that divides the
to-be-io pages to the participating devices, the previous
exofs_ios_write/read, now becomes _write/read_mirrors and a new
write/read upper layer loops on all devices calling
_write/read_mirrors. Effectively the prepare_stripes stage is the all
secret.
Also truncate need fixing to accommodate for striping.
* In a RAID0 arrangement, in a regular usage scenario, if all inode
layouts will start at the same device, the small files fill up the
first device and the later devices stay empty, the farther the device
the emptier it is.
To fix that, each inode will start at a different stripe_unit,
according to it's obj_id modulus number-of-stripe-units. And
will then span all stripe-units in the same incrementing order
wrapping back to the beginning of the device table. We call it
a stripe-units moving window.
Special consideration was taken to keep all devices in a mirror
arrangement identical. So a broken osd-device could just be cloned
from one of the mirrors and no FS scrubbing is needed. (We do that
by rotating stripe-unit at a time and not a single device at a time.)
TODO:
We no longer verify object_length == inode->i_size in exofs_iget.
(since i_size is stripped on multiple objects now).
I should introduce a multiple-device attribute reading, and use
it in exofs_iget.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
* Abstract away those members in exofs_sb_info that are related/needed
by a layout into a new exofs_layout structure. Embed it in exofs_sb_info.
* At exofs_io_state receive/keep a pointer to an exofs_layout. No need for
an exofs_sb_info pointer, all we need is at exofs_layout.
* Change any usage of above exofs_sb_info members to their new name.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
This patch changes on-disk format, it is accompanied with a parallel
patch to mkfs.exofs that enables multi-device capabilities.
After this patch, old exofs will refuse to mount a new formatted FS and
new exofs will refuse an old format. This is done by moving the magic
field offset inside the FSCB. A new FSCB *version* field was added. In
the future, exofs will refuse to mount unmatched FSCB version. To
up-grade or down-grade an exofs one must use mkfs.exofs --upgrade option
before mounting.
Introduced, a new object that contains a *device-table*. This object
contains the default *data-map* and a linear array of devices
information, which identifies the devices used in the filesystem. This
object is only written to offline by mkfs.exofs. This is why it is kept
separate from the FSCB, since the later is written to while mounted.
Same partition number, same object number is used on all devices only
the device varies.
* define the new format, then load the device table on mount time make
sure every thing is supported.
* Change I/O engine to now support Mirror IO, .i.e write same data
to multiple devices, read from a random device to spread the
read-load from multiple clients (TODO: stripe read)
Implementation notes:
A few points introduced in previous patch should be mentioned here:
* Special care was made so absolutlly all operation that have any chance
of failing are done before any osd-request is executed. This is to
minimize the need for a data consistency recovery, to only real IO
errors.
* Each IO state has a kref. It starts at 1, any osd-request executed
will increment the kref, finally when all are executed the first ref
is dropped. At IO-done, each request completion decrements the kref,
the last one to return executes the internal _last_io() routine.
_last_io() will call the registered io_state_done. On sync mode a
caller does not supply a done method, indicating a synchronous
request, the caller is put to sleep and a special io_state_done is
registered that will awaken the caller. Though also in sync mode all
operations are executed in parallel.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
In anticipation for multi-device operations, we separate osd operations
into an abstract I/O API. Currently only one device is used but later
when adding more devices, we will drive all devices in parallel according
to a "data_map" that describes how data is arranged on multiple devices.
The file system level operates, like before, as if there is one object
(inode-number) and an i_size. The io engine will split this to the same
object-number but on multiple device.
At first we introduce Mirror (raid 1) layout. But at the final outcome
we intend to fully implement the pNFS-Objects data-map, including
raid 0,4,5,6 over mirrored devices, over multiple device-groups. And
more. See: http://tools.ietf.org/html/draft-ietf-nfsv4-pnfs-obj-12
* Define an io_state based API for accessing osd storage devices
in an abstract way.
Usage:
First a caller allocates an io state with:
exofs_get_io_state(struct exofs_sb_info *sbi,
struct exofs_io_state** ios);
Then calles one of:
exofs_sbi_create(struct exofs_io_state *ios);
exofs_sbi_remove(struct exofs_io_state *ios);
exofs_sbi_write(struct exofs_io_state *ios);
exofs_sbi_read(struct exofs_io_state *ios);
exofs_oi_truncate(struct exofs_i_info *oi, u64 new_len);
And when done
exofs_put_io_state(struct exofs_io_state *ios);
* Convert all source files to use this new API
* Convert from bio_alloc to bio_kmalloc
* In io engine we make use of the now fixed osd_req_decode_sense
There are no functional changes or on disk additions after this patch.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Even though exofs has a 4k block size, statfs blocks
is in sectors (512 bytes).
Also if target returns 0 for capacity then make it
ULLONG_MAX. df does not like zero-size filesystems
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
It is important to print in the logs when a filesystem was
mounted and eventually unmounted.
Print the osd-device's osd_name and pid the FS was
mounted/unmounted on.
TODO: How to also print the namespace path the filesystem was
mounted on?
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Vladimir Davydov
David Howells
Christoph Hellwig
Boaz Harrosh
Rashika Kheria
Eric W. Biederman
Marco Stornelli
Kirill A. Shutemov
Artem Bityutskiy
Sachin Bhamare
Boaz Harrosh
Linus Torvalds
Al Viro
Dan Carpenter
Masanari Iida
Paul Gortmaker
H Hartley Sweeten
Tobias Klauser
Nick Piggin
Jens Axboe
Tejun Heo