Commit Graph

10712 Commits

Author SHA1 Message Date
Ed Schouten
f6dd5c15fd Import some improvements to the TTY code from the MPSAFE TTY branch.
- Change the ddb(4) commands to be more useful (by thompsa@):
  - `show ttys' is now called `show all ttys'. This command will now
    also display the address where the TTY data structure resides.
  - Add `show tty <addr>', which dumps the TTY in a readable form.

- Place an upper bound on the TTY buffer sizes. Some drivers do not want
  to care about baud rates. Protect these drivers by preventing the TTY
  buffers from getting enormous. Right now we'll just clamp it to 64K,
  which is pretty high, taking into account that these buffers are only
  used by the built-in discipline.

- Only call ttydev_leave() when needed. Back in April/May the TTY
  reference counting mechanism was a little different, which required us
  to call ttydev_leave() each time we finished a cdev operation.
  Nowadays we only need to call ttydev_leave() when we really mark it as
  being closed.

- Improve return codes of read() and write() on TTY device nodes.

- Make sure we really wake up all blocked threads when the driver calls
  tty_rel_gone(). There were some possible code paths where we didn't
  properly wake up any readers/writers.

- Add extra assertions to prevent sleeping on a TTY that has been
  abandoned by the driver.

- Use ttydev_cdevsw as a more reliable method to figure out whether a
  device node is a real TTY device node.

Obtained from:	//depot/projects/mpsafetty/...
Reviewed by:	thompsa
2008-10-15 16:58:35 +00:00
David Xu
904c5ec4e3 Move per-thread userland debugging flags into seperated field,
this eliminates some problems of locking, e.g, a thread lock is needed
but can not be used at that time. Only the process lock is needed now
for new field.
2008-10-15 06:31:37 +00:00
Roman Divacky
e4866772c3 Check the result of copyin and in a case of error
return one. This prevents setting wrong priority
or (more likely) returning EINVAL.

Approved by:	kib (mentor)
2008-10-13 21:04:52 +00:00
Robert Watson
ac2456bfc3 Downgrade XXX to a Note for fgetsock() and fputsock().
MFC after:	3 days
2008-10-12 20:03:17 +00:00
Robert Watson
4b0e2b9add Remove stale comment: while uipc_connect2() was, until recently, not
static so it could be used by fifofs (actually portalfs), it is now
static.

Submitted by:	kensmith
2008-10-11 17:28:22 +00:00
Attilio Rao
0d7935fd01 Remove the struct thread unuseful argument from bufobj interface.
In particular following functions KPI results modified:
- bufobj_invalbuf()
- bufsync()

and BO_SYNC() "virtual method" of the buffer objects set.
Main consumers of bufobj functions are affected by this change too and,
in particular, functions which changed their KPI are:
- vinvalbuf()
- g_vfs_close()

Due to the KPI breakage, __FreeBSD_version will be bumped in a later
commit.

As a side note, please consider just temporary the 'curthread' argument
passing to VOP_SYNC() (in bufsync()) as it will be axed out ASAP

Reviewed by:	kib
Tested by:	Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-10-10 21:23:50 +00:00
Warner Losh
53366abd02 Close, but not eliminate, a race condition. It is one that properly
designed drivers would never hit, but was exposed in diving into
another problem...

When expanding the devclass array, free the old memory after updating
the pointer to the new memory.  For the following single race case,
this helps:

	allocate new memory
	copy to new memory
	free old memory
<interrupt>				read pointer to freed memory
	update pointer to new memory

Now we do
	allocate new memory
	copy to new memory
	update pointer to new memory
	free old memory

Which closes this problem, but doesn't even begin to address the
multicpu races, which all should be covered by Giant at the moment,
but likely aren't completely.

Note: reviewers were ok with this fix, but suggested the use case
wasn't one we wanted to encourage.

Reviewed by:	jhb, scottl.
2008-10-10 17:49:47 +00:00
Konstantin Belousov
387ad99800 If the ABI-overriden interpreter was not loaded, do not set
have_interp to TRUE. This allows the code in image activator to try
/libexec/ld-elf.so.1 as interpreter when newinterp is not found to
execute.

Reviewed by:	peter
MFC after:	2 weeks (together with r175105)
2008-10-08 11:11:36 +00:00
Robert Watson
e298cf5902 Remove stale comment (and XXX saying so) about why we zero the file
descriptor pointer in unp_freerights: we can no longer recurse into
unp_gc due to unp_gc being invoked in a deferred way, but it's still
a good idea.

MFC after:	3 days
2008-10-08 06:26:51 +00:00
Robert Watson
fa9402f28a Differentiate pr_usrreqs for stream and datagram UNIX domain sockets, and
employ soreceive_dgram for the datagram case.

MFC after:	3 months
2008-10-08 06:19:49 +00:00
Robert Watson
ff601c3645 In soreceive_dgram, when a 0-length buffer is passed into recv(2) and
no data is ready, return 0 rather than blocking or returning EAGAIN.
This is consistent with the behavior of soreceive_generic (soreceive)
in earlier versions of FreeBSD, and restores this behavior for UDP.

Discussed with:	jhb, sam
MFC after:	3 days
2008-10-07 20:57:55 +00:00
Robert Watson
ffe72750d9 Remove temporary debugging KASSERT's introduced to detect protocols
improperly invoking sosend(), soreceive(), and sopoll() instead of
attach either specialized or _generic() versions of those functions
to their pru_sosend, pru_soreceive, and pru_sopoll protosw methods.

MFC after:	3 days
2008-10-07 09:57:03 +00:00
Robert Watson
7978014d3a Rewrite sbreserve_locked()'s comment on NULL thread pointers, eliminating
an XXXRW about the comment being stale.

MFC after:	3 days
2008-10-07 09:51:39 +00:00
Robert Watson
58f7ce962c Lock receive socket buffer in soo_stat() rather than commenting that we
should lock it, which may marginally improve the consistency of the
results.  Remove comment.

MFC after:	3 days
2008-10-07 07:10:28 +00:00
Robert Watson
2c8995842c Now that portalfs doesn't directly invoke uipc_connect2(), make it a
static symbol.

MFC after:	3 days
2008-10-06 18:43:11 +00:00
Sam Leffler
73254c9ee7 dynamically allocate the task structure in firmware_mountroot: when
booting from an MFS root (e.g. from an install CD) firmware_mountroot
can be called twice with the second call happening before the task
callback occurs; this results in the task structure contents being
corrupted because it was declared static.

Submitted by:	marius (original version)
2008-10-04 23:58:02 +00:00
John Baldwin
4ec9dc4de2 Oops, missed updating a place with with 's/lock1/plock/' when adding
interlock support to WITNESS.  Specifically, the printf listing the
first location when duplicate locks of the same type are acquired.

Reported by:	pho
2008-10-03 18:13:05 +00:00
Robert Watson
0b36cd25fc Further minor cleanups to UNIX domain sockets:
- Staticize and locally prototype functions uipc_ctloutput(), unp_dispose(),
  unp_init(), and unp_externalize(), none of which have been required
  outside of uipc_usrreq.c since uipc_proto.c was removed.
- Remove stale prototype for uipc_usrreq(), which has not existed in the
  code since 1997
- Forward declare and staticize uipc_usrreqs structure in uipc_usrreq.c and
  not un.h.
- Comment on why uipc_connect2() is still non-static -- it is used directly
  by fifofs.
- Remove stale comments, tidy up whitespace.

MFC after:	3 days (where applicable)
2008-10-03 13:01:56 +00:00
Robert Watson
60a5ef26a1 Remove or update several stale comments.
A bit of whitespace/style cleanup.

Update copyright.

MFC after:	3 days (applicable changes)
2008-10-03 09:01:55 +00:00
Marko Zec
8b615593fc Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit

Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.

Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().

Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).

All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).

(*) netipsec/keysock.c did not validate depending on compile time options.

Implemented by:	julian, bz, brooks, zec
Reviewed by:	julian, bz, brooks, kris, rwatson, ...
Approved by:	julian (mentor)
Obtained from:	//depot/projects/vimage-commit2/...
X-MFC after:	never
Sponsored by:	NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
Peter Wemm
e6592ee55c Collect N identical (or near identical) mkdumpheader() implementations into
one, as threatened in the comment.  Textdump magic can be passed in.
2008-10-01 22:08:53 +00:00
John Baldwin
b957a8225a Enable shared locks for path name lookups on supported filesystems (NFS
client, UFS, and ZFS) by default.
2008-10-01 19:25:37 +00:00
John Baldwin
d59701d07d Remove the LOOKUP_SHARED kernel option. Instead, make vfs.lookup_shared
a loader tunable (it was already a sysctl).
2008-10-01 19:24:16 +00:00
John Baldwin
1af1c6cd8a Wait until after dropping the receive socket buffer lock to allocate space
to store the socket address stored in the first mbuf in a packet chain.
This reduces contention on the lock and CPU system time in certain UDP
workloads.

Tested by:	ps
Reviewed by:	rwatson
MFC after:	1 week
2008-10-01 19:14:05 +00:00
Robert Watson
25edc6dd20 Various cleanups for soreceive_dgram():
- Update or remove comments that were left over from the original
  soreceive_generic() implementation.  Quite a few were misleading in the
  context of the new code.
- Since soreceive_dgram() has a simpler structure, replace several gotos
  with a while loop making the invariants more clear.
- In the blocking while loop, don't try to handle cases incompatible with
  the loop invariant (since m is always NULL, don't check for and handle
  non-NULL).
- Don't drop and re-acquire the socket buffer lock unnecessarily after
  sbwait() returns, which may help reduce lock contention (etc).
- Assume PR_ATOMIC since we assert it at the top of the function.

MFC after:	3 days
2008-10-01 13:26:52 +00:00
John Baldwin
c4688866d3 Update the function name in several assertions in soreceive_dgram().
Approved by:	rwatson
MFC after:	3 days
2008-09-30 18:44:26 +00:00
Konstantin Belousov
41a4e90e6f If the panic thread is preempted after setting panicstr but before
setting TDF_INPANIC then it will never be rescheduled again. Wrap
setting the panic condition with the critical section.

Noted and reviewed by:	tegge
MFC after:	1 week
2008-09-27 15:45:54 +00:00
Ed Schouten
c6ec8c53c5 Move uminor() and umajor() to the same place as userspace minor() and major().
The uminor() and umajor() functions have the same use in kernel space as
the minor() and major() functions in userspace. If we ever get rid of
the minor() function in kernel space, we could decide to just expose
minor() and major() to kernel space, making uminor() and umajor()
redundant.

There are two reasons why we want to have uminor() and umajor() in
<sys/types.h>:

- Having them close together prevents them from diverting. Even though
  it's unlikely the definitions will change, it's a good habit to have
  them at the same place.

- They don't really belong in kern_conf.c. kern_conf.c has been
  liberated from dealing with device major and minor number handling.

The device_ids(9) manpage now lists the wrong #include's, because it
should only list <sys/types.h> now. I'm leaving it as it is now, because
I wonder if we should document them anyway. We're probably better off
documenting minor(3) and major(3).
2008-09-27 13:19:09 +00:00
Ed Schouten
6bfa9a2d66 Replace all calls to minor() with dev2unit().
After I removed all the unit2minor()/minor2unit() calls from the kernel
yesterday, I realised calling minor() everywhere is quite confusing.
Character devices now only have the ability to store a unit number, not
a minor number. Remove the confusion by using dev2unit() everywhere.

This commit could also be considered as a bug fix. A lot of drivers call
minor(), while they should actually be calling dev2unit(). In -CURRENT
this isn't a problem, but it turns out we never had any problem reports
related to that issue in the past. I suspect not many people connect
more than 256 pieces of the same hardware.

Reviewed by:	kib
2008-09-27 08:51:18 +00:00
Ed Schouten
dacf7de1a6 Don't forget to initialize `int error' in ttydev_open().
I've had some reports in the past that opening an already opened TTY
through, for example, /dev/tty can fail with random error codes. Looking
at ttydev_open(), I can see there is a way `error' is returned without
initialising it. Even though I haven't had any confirmation this fixes
the bug, I'll fix it anyway.

Reported by:	Andrzej Tobola <ato iem pw edu pl>
2008-09-26 18:17:04 +00:00
Ed Schouten
edde874555 Rename the minor' argument of make_dev(9) to unit'.
To prevent any further confusion about device minor and unit numbers,
we'd better just refer to device unit numbers. Many people still think
the numbers we show inside devfs have any relation to the numbers passed
to make_dev(9), which is not the case.

Discussed with:	kib
2008-09-26 14:31:24 +00:00
Ed Schouten
d3ce832719 Remove unit2minor() use from kernel code.
When I changed kern_conf.c three months ago I made device unit numbers
equal to (unneeded) device minor numbers. We used to require
bitshifting, because there were eight bits in the middle that were
reserved for a device major number. Not very long after I turned
dev2unit(), minor(), unit2minor() and minor2unit() into macro's.
The unit2minor() and minor2unit() macro's were no-ops.

We'd better not remove these four macro's from the kernel, because there
is a lot of (external) code that may still depend on them. For now it's
harmless to remove all invocations of unit2minor() and minor2unit().

Reviewed by:	kib
2008-09-26 14:19:52 +00:00
John Baldwin
74d9b5a551 Regen. 2008-09-25 20:08:36 +00:00
John Baldwin
48a43ae819 Tidy up a few things with syscall generation:
- Instead of using a syscall slot (370) just to get a function prototype
  for lkmressys(), add an explicit function prototype to <sys/sysent.h>.
  This also removes unused special case checks for 'lkmressys' from
  makesyscalls.sh.
- Instead of having magic logic in makesyscalls.sh to only generate a
  function prototype the first time 'lkmnosys' is seen, make 'NODEF'
  always not generate a function prototype and include an explicit
  prototype for 'lkmnosys' in <sys/sysent.h>.
- As a result of the fix in (2), update the LKM syscall entries in
  the freebsd32 syscall table to use 'lkmnosys' rather than 'nosys'.
- Use NOPROTO for the __syscall() entry (198) in the native ABI.  This
  avoids the need for magic logic in makesyscalls.h to only generate
  a function prototype the first time 'nosys' is encountered.
2008-09-25 20:07:42 +00:00
John Baldwin
7d43ca696e - Don't do a WITNESS_SAVE() on the interlock if it is Giant in the condition
variable wait routines.  DROP_GIANT() already manages that state in the
  Giant interlock case.
- Assert that Giant is held when it is passed as a sleep interlock.
2008-09-25 13:42:19 +00:00
John Baldwin
d2722d704c Part 1 of making shared lookups more resilient with respect to forced
unmounts.  When we upgrade a vnode lock from shared to exclusive during
a name cache lookup, fail the lookup with EBADF if the vnode is invalidated
while we are waiting for the exclusive lock.

Also, for correctness (though I'm not sure it can occur in practice),
downgrade an exclusively locked vnode if it should be share locked.

Tested by:	pho
2008-09-24 18:51:33 +00:00
John Baldwin
c1fa2e4200 Update description of witness_watch. 2008-09-24 18:47:24 +00:00
Ed Schouten
4c7428e1ff Fix a crash when calling tty_rel_free() while draining during closure.
Yesterday I got two reports of potential crashes, related to TTY
deallocation during device closure. When a thread is in TF_OPENCLOSE,
draining its output upon closure, we should not allow calls to
tty_rel_free() to happen at the same time. This could cause the TTY to
be torn down twice.

PR:		kern/127561
Reported by:	KOIE Hidetaka <koie suri co jp>
Discussed with:	thompsa
2008-09-24 11:16:09 +00:00
Konstantin Belousov
a8d403e102 Change the static struct sysentvec and struct Elf_Brandinfo initializers
to the C99 style. At least, it is easier to read sysent definitions
that way, and search for the actual instances of sigcode etc.

Explicitely initialize sysentvec.sv_maxssiz that was missed in most
sysvecs.

No objection from:	jhb
MFC after:	1 month
2008-09-24 10:14:37 +00:00
Ed Schouten
b61637107c Track state to determine if the associated TTY device node has been used.
It turns out our old TTY layer (and other implementations) block when
you read() on a PTY master device of which the slave device node has not
been opened yet. Our new implementation just returned 0. This caused
applications like telnetd to die in a very subtle way (when child
processes would open the TTY later than the first call to select()).

Introduce a new flag called PTS_FINISHED, which indicates whether we
should block or bail out of a read() or write() occurs.

Reported by:	Claude Buisson <clbuisson orange fr>
2008-09-23 17:12:25 +00:00
David E. O'Brien
b386eb9139 style(9) 2008-09-23 14:25:56 +00:00
David E. O'Brien
715457f6f6 Reverse if() logic to improve readability.
Reviewed by:	ru
2008-09-23 14:25:38 +00:00
Ed Schouten
a1215e37a4 Introduce a hooks layer for the MPSAFE TTY layer.
One of the features that prevented us from fixing some of the TTY
consumers to work once again, was an interface that allowed consumers to
do the following:

- `Sniff' incoming data, which is used by the snp(4) driver.

- Take direct control of the input and output paths of a TTY, which is
  used by ng_tty(4), ppp(4), sl(4), etc.

There's no practical advantage in committing a hooks layer without
having any consumers. In P4 there is a preliminary port of snp(4) and
thompsa@ is busy porting ng_tty(4) to this interface. I already want to
have it in the tree, because this may stimulate others to work on the
remaining modules.

Discussed with:	thompsa
Obtained from:	//depot/projects/mpsafetty/...
2008-09-22 19:25:14 +00:00
Ed Schouten
d344ffe549 Fix style(9) issue in TTY header files: document function argument names.
According to style(9), function argument names should only be omitted
for prototypes that are exported to userspace. This means we should
document the function arguments in the TTY header files, because they
are only used in userspace.

While there, change the type of the buffer argument of
ttydisc_rint_bypass() to `const void *' instead of `char *'.

Requested by:	attilio
Obtained from:	//depot/projects/mpsafetty/...
2008-09-22 18:44:09 +00:00
Joseph Koshy
122ccdc1ca Support sparsely numbered CPUs.
Requested by:	obrien, alfred (long ago)
2008-09-22 10:37:02 +00:00
Ed Schouten
37ddf38e38 Make fstat() on a pseudo-terminal master return sane timestamps.
Because pseudo-terminal master file descriptors no longer have a vnode
underneath, we have to fill in fstat() values ourselves. Make our
implementation somewhat sane by returning the timestamps of the TTY
device node that corresponds with our file descriptor.

Obtained from:	//depot/projects/mpsafettty/...
2008-09-21 19:24:15 +00:00
Ed Schouten
3111c5c922 Now that the number of clist consumers have dropped massively, trim down
the code to prevent useless waste of space.

- Remove support for quote bits. There is not a single driver that needs
  these bits anymore. This means putc() now accepts a char instead of an
  int.

- Remove the unneeded catq() and nextc() routines. They were only used
  by the old TTY layer.

- Convert the clist code to use ANSI C prototypes.
2008-09-21 18:12:18 +00:00
Konstantin Belousov
caf8aec886 fdescfs, devfs, mqueuefs, nfs, portalfs, pseudofs, tmpfs and xfs
initialize the vattr structure in VOP_GETATTR() with VATTR_NULL(),
vattr_null() or by zeroing it. Remove these to allow preinitialization
of fields work in vn_stat(). This is needed to get birthtime initialized
correctly.

Submitted by:   Jaakko Heinonen <jh saunalahti fi>
Discussed on:   freebsd-fs
MFC after:	1 month
2008-09-20 19:50:52 +00:00
Konstantin Belousov
4c5a20e3da Initialize va_rdev to NODEV instead of 0 or VNOVAL in VOP_GETATTR().
NODEV is more appropriate when va_rdev doesn't have a meaningful value.

Submitted by:   Jaakko Heinonen <jh saunalahti fi>
Suggested by:   bde
Discussed on:   freebsd-fs
MFC after:	1 month
2008-09-20 19:49:15 +00:00
Konstantin Belousov
0fbbf2ea56 Initialize va_rdev to NODEV and va_fsid to VNOVAL before the
VOP_GETATTR() call in vn_stat(). Thus if a file system doesn't
initialize those fields in VOP_GETATTR() they will have a sane default
value.

Submitted by:   Jaakko Heinonen <jh saunalahti fi>
Discussed on:   freebsd-fs
MFC after:	1 month
2008-09-20 19:48:24 +00:00