09 Jan, 2006
40 commits
-
Some long time ago, dentry struct was carefully tuned so that on 32 bits
UP, sizeof(struct dentry) was exactly 128, ie a power of 2, and a multiple
of memory cache lines.Then RCU was added and dentry struct enlarged by two pointers, with nice
results for SMP, but not so good on UP, because breaking the above tuning
(128 + 8 = 136 bytes)This patch reverts this unwanted side effect, by using an union (d_u),
where d_rcu and d_child are placed so that these two fields can share their
memory needs.At the time d_free() is called (and d_rcu is really used), d_child is known
to be empty and not touched by the dentry freeing.Lockless lookups only access d_name, d_parent, d_lock, d_op, d_flags (so
the previous content of d_child is not needed if said dentry was unhashed
but still accessed by a CPU because of RCU constraints)As dentry cache easily contains millions of entries, a size reduction is
worth the extra complexity of the ugly C union.Signed-off-by: Eric Dumazet
Cc: Dipankar Sarma
Cc: Maneesh Soni
Cc: Miklos Szeredi
Cc: "Paul E. McKenney"
Cc: Ian Kent
Cc: Paul Jackson
Cc: Al Viro
Cc: Christoph Hellwig
Cc: Trond Myklebust
Cc: Neil Brown
Cc: James Morris
Cc: Stephen Smalley
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
It would be helpful if the kernel did not silently stop parsing
nfs options, but instead warned about any he does not recognize. The
attached patch adds one printk to do just that.It took me a couple of hours to find my configuration mistake.
Cc: "David S. Miller"
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The only user of send_sigio_to_task() already holds tasklist_lock, so it is
better not to send the signal via send_group_sig_info() (which takes
tasklist recursively) but use group_send_sig_info().The same change in send_sigurg()->send_sigurg_to_task().
Signed-off-by: Oleg Nesterov
Cc: "Paul E. McKenney"
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch removes unneeded sig->curr_target recalculation under 'if
(atomic_dec_and_test(&sig->count))' in __exit_signal().When sig->count == 0 the signal can't be sent to this task and
next_thread(tsk) == tsk anyway.Signed-off-by: Oleg Nesterov
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
uniq -d MAINTAINERS
Signed-off-by: Nicolas Kaiser
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
There are places in the resize code in which EXT3_SB() macro is used after
an statement like sbi = EXT3_SB(sb) is done. Inside the same function,
both sbi and EXT3_SB() are used to reference the super block Altough it is
not wrong, keeping it coherent increases legibility, IMHO.Signed-off-by: Glauber de Oliveira Costa
Cc: "Stephen C. Tweedie"
Cc: Andreas Dilger
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Remove the trailing newlines in calls to ext3_warning(). This function
already adds a trailing newline to the end of messages.Signed-off-by: Glauber de Oliveira Costa
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Make oprofile alloc_cpu_buffers() function NUMA aware, allocating each CPU
local buffer in its memory node if possible.Signed-off-by: Eric Dumazet
Cc: Philippe Elie
Cc: John Levon
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The patch below adds a new mount option to allow the external journal
device to be specified.The syntax is as follows:
# mount -t ext3 -o journal_dev=0x0820 ...
where 0x0820 means major=8 and minor=32.Signed-off-by: Johann Lombardi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Small cleanups in shared mounts code.
Signed-off-by: Miklos Szeredi
Cc: Ram Pai
Cc:
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Neil Brown
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Thanks to Nathan Lynch for the review and comments. Thanks to Joel Schopp
for the pointer to add user space scipts.Signed-off-by: Ashok Raj
Signed-off-by: Nathan Lynch
Signed-off-by: Joel Schopp
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
According to the TCG specifications measurements or hashes of the BIOS code
and data are extended into TPM PCRS and a log is kept in an ACPI table of
these extensions for later validation if desired. This patch exports the
values in the ACPI table through a security-fs seq_file.Signed-off-by: Seiji Munetoh
Signed-off-by: Stefan Berger
Signed-off-by: Reiner Sailer
Signed-off-by: Kylene Hall
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
zap_other_threads() sets SIGNAL_GROUP_EXIT at the very start,
do_group_exit() doesn't need to do it.Signed-off-by: Oleg Nesterov
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
__group_complete_signal() sets ->group_stop_count in sig_kernel_coredump()
path and marks the target thread as ->group_exit_task. So any thread
except group_exit_task will go to handle_group_stop()->finish_stop().However, when group_exit_task actually starts do_coredump(), it sets
SIGNAL_GROUP_EXIT, but does not reset ->group_stop_count while killing
other threads. If we have not yet stopped threads in the same thread
group, they all will spin in kernel mode until group_exit_task sends them
SIGKILL, because ->group_stop_count > 0 means:recalc_sigpending_tsk() never clears TIF_SIGPENDING
get_signal_to_deliver() goes to handle_group_stop()
handle_group_stop() returns when SIGNAL_GROUP_EXIT set
syscall_exit/resume_userspace notice TIF_SIGPENDING,
call get_signal_to_deliver() again.So we are wasting cpu cycles, and if one of these threads is rt_task() this
may be a serious problem.NOTE: do_coredump() holds ->mmap_sem, so not stopped threads can't escape
coredumping after clearing ->group_stop_count.See also this thread: http://marc.theaimsgroup.com/?t=112739139900002
Signed-off-by: Oleg Nesterov
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Use symbolic names instead of hardcoded constants.
Signed-off-by: Oleg Nesterov
Acked-by: Harald Welte
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
We've had two instances recently of overflows when doing
64_bit_value = (32_bit_value << PAGE_CACHE_SHIFT)
I did a tree-wide grep of `<page_base)
Cc: Oleg Drokin
Cc: David Howells
Cc: David Woodhouse
Cc:
Cc: Christoph Hellwig
Cc: Anton Altaparmakov
Cc: Jeff Dike
Cc: Paolo 'Blaisorblade' Giarrusso
Cc: Roman Zippel
Cc:
Cc: Miklos Szeredi
Cc: Russell King
Cc: Trond Myklebust
Cc: Neil Brown
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
__create_workqueue() not checking return of alloc_percpu()
NULL dereference was possible.
Signed-off-by: Ben Collins
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Sent by Paul Clements , who needs to read
Documentation/SubmittingPatches..Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
HDIO_GETGEO is implemented in most block drivers, and all of them have to
duplicate the code to copy the structure to userspace, as well as getting
the start sector. This patch moves that to common code [1] and adds a
->getgeo method to fill out the raw kernel hd_geometry structure. For many
drivers this means ->ioctl can go away now.[1] the s390 block drivers are odd in this respect. xpram sets ->start
to 4 always which seems more than odd, and the dasd driver shifts
the start offset around, probably because of it's non-standard
sector size.Signed-off-by: Christoph Hellwig
Cc: Jens Axboe
Cc:
Cc: Jeff Dike
Cc: Paolo Giarrusso
Cc: Bartlomiej Zolnierkiewicz
Cc: Neil Brown
Cc: Markus Lidel
Cc: Russell King
Cc: David Woodhouse
Cc: Martin Schwidefsky
Cc: James Bottomley
Signed-off-by: Adrian Bunk
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
While rooting aroung in the signal code trying to understand how to fix the
SIG_IGN ploy (set sig handler to SIG_IGN and flood system with high speed
repeating timers) I came across what, I think, is a problem in sigaction()
in that when processing a SIG_IGN request it flushes signals from 1 to
SIGRTMIN and leaves the rest. Attempt to fix this.Signed-off-by: George Anzinger
Cc: Roland McGrath
Cc: Linus Torvalds
Signed-off-by: Adrian Bunk
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Make it possible for a running process (such as gssapid) to be able to
instantiate a key, as was requested by Trond Myklebust for NFS4.The patch makes the following changes:
(1) A new, optional key type method has been added. This permits a key type
to intercept requests at the point /sbin/request-key is about to be
spawned and do something else with them - passing them over the
rpc_pipefs files or netlink sockets for instance.The uninstantiated key, the authorisation key and the intended operation
name are passed to the method.(2) The callout_info is no longer passed as an argument to /sbin/request-key
to prevent unauthorised viewing of this data using ps or by looking in
/proc/pid/cmdline.This means that the old /sbin/request-key program will not work with the
patched kernel as it will expect to see an extra argument that is no
longer there.A revised keyutils package will be made available tomorrow.
(3) The callout_info is now attached to the authorisation key. Reading this
key will retrieve the information.(4) A new field has been added to the task_struct. This holds the
authorisation key currently active for a thread. Searches now look here
for the caller's set of keys rather than looking for an auth key in the
lowest level of the session keyring.This permits a thread to be servicing multiple requests at once and to
switch between them. Note that this is per-thread, not per-process, and
so is usable in multithreaded programs.The setting of this field is inherited across fork and exec.
(5) A new keyctl function (KEYCTL_ASSUME_AUTHORITY) has been added that
permits a thread to assume the authority to deal with an uninstantiated
key. Assumption is only permitted if the authorisation key associated
with the uninstantiated key is somewhere in the thread's keyrings.This function can also clear the assumption.
(6) A new magic key specifier has been added to refer to the currently
assumed authorisation key (KEY_SPEC_REQKEY_AUTH_KEY).(7) Instantiation will only proceed if the appropriate authorisation key is
assumed first. The assumed authorisation key is discarded if
instantiation is successful.(8) key_validate() is moved from the file of request_key functions to the
file of permissions functions.(9) The documentation is updated.
From:
Build fix.
Signed-off-by: David Howells
Cc: Trond Myklebust
Cc: Alexander Zangerl
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Cause any links within a keyring to keys that match a key to be linked into
that keyring to be discarded as a link to the new key is added. The match is
contingent on the type and description strings being the same.This permits requests, adds and searches to displace negative, expired,
revoked and dead keys easily. After some discussion it was concluded that
duplicate valid keys should probably be discarded also as they would otherwise
hide the new key.Since request_key() is intended to be the primary method by which keys are
added to a keyring, duplicate valid keys wouldn't be an issue there as that
function would return an existing match in preference to creating a new key.Signed-off-by: David Howells
Cc: Trond Myklebust
Cc: Alexander Zangerl
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Add a new keyctl function that allows the expiry time to be set on a key or
removed from a key, provided the caller has attribute modification access.Signed-off-by: David Howells
Cc: Trond Myklebust
Cc: Alexander Zangerl
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
kmsg_write returns with printk, so some programs may be confused by a
successful write() with a return value different than the buffer length.# /bin/echo something > /dev/kmsg
/bin/echo: write error: Inappropriate ioctl for deviceThe drawbacks is that the printk return value can no more be quickly
checked from userspace.Signed-off-by: Guillaume Chazarain
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
What's the true meaning of the printk return value? Should it include the
priority prefix length of 3? and what about the timing information? In
both cases it was broken:strace -e write echo 1 > /dev/kmsg
=> write(1, "1\n", 2) = 5
strace -e write echo "1" > /dev/kmsg
=> write(1, "1\n", 5) = 8The returned length was "length of input string + 3", I made it "length
of string output to the log buffer".Note that I couldn't find any printk caller in the kernel interested by its
return value besides kmsg_write.Signed-off-by: Guillaume Chazarain
Acked-By: Tim Bird
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
When making an fctl locking call through compat_sys_fcntl64 (i.e. a 32bit
app on a 64bit kernel), the syscall can return a locking range that is in
conflict with the queried lock.If some aspect of this range does not fit in the 32bit structure, something
needs to be done.The current code is wrong in several respects:
- It returns data to userspace even if no conflict was found
i.e. it should check l_type for F_UNLCK
- It returns -EOVERFLOW too agressively. A lock range covering
the last possible byte of the file (start = COMPAT_OFF_T_MAX,
len = 1) should be possible, but is rejected with the current test.
- A extra-long 'len' should not be a problem. If only that part
of the conflicting lock that would be visible to the 32bit
app needs to be reported to the 32bit app anyway.This patch addresses those three issues and adds a comment to (hopefully)
record it for posterity.Note: this patch mainly affects test-cases. Real applications rarely is
ever see the problems.This patch has been tested (LSB test suite), and works.
Signed-off-by: Neil Brown
Cc: Arnd Bergmann
Cc: Christoph Hellwig
Cc: Matthew Wilcox
Cc: Trond Myklebust
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
SUS requires that when truncating a file to the size that it currently
is:
truncate and ftruncate should NOT modify ctime or mtime
O_TRUNC SHOULD modify ctime and mtime.Currently mtime and ctime are always modified on most local
filesystems (side effect of ->truncate) or never modified (on NFS).With this patch:
ATTR_CTIME|ATTR_MTIME are sent with ATTR_SIZE precisely when
an update of these times is required whether size changes or not
(via a new argument to do_truncate). This allows NFS to do
the right thing for O_TRUNC.
inode_setattr nolonger forces ATTR_MTIME|ATTR_CTIME when the ATTR_SIZE
sets the size to it's current value. This allows local filesystems
to do the right thing for f?truncate.Also, the logic in inode_setattr is changed a bit so there are two return
points. One returns the error from vmtruncate if it failed, the other
returns 0 (there can be no other failure).Finally, if vmtruncate succeeds, and ATTR_SIZE is the only change
requested, we now fall-through and mark_inode_dirty. If a filesystem did
not have a ->truncate function, then vmtruncate will have changed i_size,
without marking the inode as 'dirty', and I think this is wrong.Signed-off-by: Neil Brown
Cc: Christoph Hellwig
Cc: Trond Myklebust
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Make it possible to include linux/pagevec.h multiple times without
incurring errors due to duplicate definitions.Signed-off-by: David Howells
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Reported from Redhat Bugzilla Bug 170450
"I updated to the development kernel and now during boot only the top of the
text is visable. For example the monitor screen the is the lines and I can
only see text in the asterisk area. -
When doublescan mode is in use, scanlines must be doubled.
Thanks to Jason Dravet for reporting and testing.
Signed-off-by: Samuel Thibault
Cc:
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
inode can never be NULL when calling this function.
Signed-off-by: Christoph Hellwig
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The ptrace_get_task_struct() helper that I added as part of the ptrace
consolidation is useful in variety of places that currently opencode it.
Switch them to the common helpers.Add a ptrace_traceme() helper that needs to be explicitly called, and simplify
the ptrace_get_task_struct() interface. We don't need the request argument
now, and we return the task_struct directly, using ERR_PTR() for error
returns. It's a bit more code in the callers, but we have two sane routines
that do one thing well now.Signed-off-by: Christoph Hellwig
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
librelay and relay-app.h have been retired - update Documentation to reflect
that.Signed-off-by: Tom Zanussi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch renames relayfs_file_operations to relay_file_operations, and the
file operations themselves from relayfs_XXX to relay_file_XXX, to make it more
clear that they refer to relay files.Signed-off-by: Tom Zanussi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Documentation update for creating global buffers.
Signed-off-by: Tom Zanussi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch adds the optional is_global outparam to the create_buf_file()
callback. This can be used by clients to create a single global relayfs
buffer instead of the default per-cpu buffers. This was suggested as being
useful for certain debugging applications where it's more convenient to be
able to get all the data from a single channel without having to go to the
bother of dealing with per-cpu files.Signed-off-by: Tom Zanussi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Documentation update for creating relay files in other filesystems.
Signed-off-by: Tom Zanussi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch adds a couple of callback functions that allow a client to hook
into relay_open()/close() and supply the files that will be used to represent
the channel buffers; the default implementation if no callbacks are defined is
to create the files in relayfs. This is to support the creation and use of
relay files in other filesystems such as debugfs, as implied by the fact that
relayfs_file_operations are exported.Signed-off-by: Tom Zanussi
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
