- 01 Sep, 2009 9 commits
-
-
Christoph Hellwig authored
xfs_trans_iget is a wrapper for xfs_iget that adds the inode to the transaction after it is read. Except when the inode already is in the inode cache, in which case it returns the existing locked inode with increment lock recursion counts. Now, no one in the tree every decrements these lock recursion counts, so any user of this gets a potential double unlock when both the original owner of the inode and the xfs_trans_iget caller unlock it. When looking back in a git bisect in the historic XFS tree there was only one place that decremented these counts, xfs_trans_iput. Introduced in commit ca25df7a840f426eb566d52667b6950b92bb84b5 by Adam Sweeney in 1993, and removed in commit 19f899a3ab155ff6a49c0c79b06f2f61059afaf3 by Steve Lord in 2003. And as long as it didn't slip through git bisects cracks never actually used in that time frame. A quick audit of the callers of xfs_trans_iget shows that no caller really relies on this behaviour fortunately - xfs_ialloc allows this inode from disk so it must not be there before, and all the RT allocator routines only every add each RT bitmap inode once. In addition to removing lots of code and reducing the size of the inode item this patch also avoids the double inode cache lookup in each create/mkdir/mknod transaction. Signed-off-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
Alex Elder <aelder@sgi.com> Signed-off-by:
Felix Blyakher <felixb@sgi.com>
-
Christoph Hellwig authored
The guarantees for O_SYNC are exactly the same as the ones we need to make for an fsync call (and given that Linux O_SYNC is O_DSYNC the equivalent is fdadatasync, but we treat both the same in XFS), except with a range data writeout. Jan Kara has started unifying these two path for filesystems using the generic helpers, and I've started to look at XFS. The actual transaction commited by xfs_fsync and xfs_write_sync_logforce has a different transaction number, but actually is exactly the same. We'll only use the fsync transaction going forward. One major difference is that xfs_write_sync_logforce never issues a cache flush unless we commit a transaction causing that as a side-effect, which is an obvious bug in the O_SYNC handling. Second all the locking and i_update_size vs i_update_core changes from 978b7237 never made it to xfs_write_sync_logforce, so we add them back. To make xfs_fsync easily usable from the O_SYNC path, the filemap_fdatawait call is moved up to xfs_file_fsync, so that we don't wait on the whole file after we already waited for our portion in xfs_write. We'll also use a plain call to filemap_write_and_wait_range instead of the previous sync_page_rang which did it in two steps including an half-hearted inode write out that doesn't help us. Once we're done with this also remove the now useless i_update_size tracking. Signed-off-by:
-
Dave Chinner authored
Don't search too far - abort if it is outside a certain radius and simply do a linear search for the first free inode. In AGs with a million inodes this can speed up allocation speed by 3-4x. [hch: ported to the new xfs_ialloc.c world order] Signed-off-by:
Dave Chinner <dgc@sgi.com> Signed-off-by:
-
Christoph Hellwig authored
Currenly we have a xfs_inobt_lookup* variant for each comparism direction, and all these get all three fields of the inobt records passed, while the common case is just looking for the inode number and we have only marginally more callers than xfs_inobt_lookup* variants. So opencode a direct call to xfs_btree_lookup for the single case where we need all fields, and replace xfs_inobt_lookup* with a xfs_inobt_looku that just takes the inode number and the direction for all other callers. Signed-off-by:
-
Christoph Hellwig authored
Clarify the control flow in xfs_dialloc. Factor out a helper to go to the next node from the current one and improve the control flow by expanding composite if statements and using gotos. The xfs_ialloc_next_rec helper is borrowed from Dave Chinners dynamic allocation policy patches. Signed-off-by:
-
Dave Chinner authored
Factor out a common helper from repeated debug checks in xfs_dialloc and xfs_difree. [hch: split out from Dave's dynamic allocation policy patches] Signed-off-by:
-
Christoph Hellwig authored
Both callers of xfs_inobt_update have the record in form of a xfs_inobt_rec_incore_t, so just pass a pointer to it instead of the individual variables. Signed-off-by:
-
Christoph Hellwig authored
Most callers of xfs_inobt_get_rec need to fill a xfs_inobt_rec_incore_t, and those who don't yet are fine with a xfs_inobt_rec_incore_t, instead of the three individual variables, too. So just change xfs_inobt_get_rec to write the output into a xfs_inobt_rec_incore_t directly. Signed-off-by:
-
Dave Chinner authored
Factor out code to initialize new inode clusters into a function of it's own. This keeps xfs_ialloc_ag_alloc smaller and better structured and enables a future inode cluster initialization transaction. Also initialize the agno variable earlier in xfs_ialloc_ag_alloc to avoid repeated byte swaps. [hch: The original patch is from Dave from his unpublished inode create transaction patch series, with some modifcations by me to apply stand-alone] Signed-off-by:
Dave Chinner <david@fromorbit.com> Signed-off-by:
-
- 31 Aug, 2009 5 commits
-
-
-
Julia Lawall authored
bp was tested for NULL a few lines before, followed by a return, and there is no intervening modification of its value. A simplified version of the semantic match that finds this problem is as follows: (http://www.emn.fr/x-info/coccinelle/) // <smpl> @r exists@ local idexpression x; expression E; position p1,p2; @@ if (x == NULL || ...) { ... when forall return ...; } ... when != \(x=E\|x--\|x++\|--x\|++x\|x-=E\|x+=E\|x|=E\|x&=E\|&x\) ( *x == NULL | *x != NULL ) // </smpl> Signed-off-by:
Julia Lawall <julia@diku.dk> Acked-by:
-
Eric Sandeen authored
Commit a19d9f88 removed the ino64 option but left the XFS_INO64_OFFSET define it used in place - just remove it. Signed-off-by:
Eric Sandeen <sandeen@sandeen.net> Reviewed-by:
-
Eric Sandeen authored
CONFIG_XFS_DEBUG builds still need xfs_read_agf to be non-static, oops. Signed-off-by:
-
Eric Sandeen authored
A lot more functions could be made static, but they need forward declarations; this does some easy ones, and also found a few unused functions in the process. Signed-off-by:
-
- 18 Aug, 2009 1 commit
-
-
Christoph Hellwig authored
The locking in xfs_iget_cache_hit currently has numerous problems: - we clear the reclaim tag without i_flags_lock which protects modifications to it - we call inode_init_always which can sleep with pag_ici_lock held (this is oss.sgi.com BZ #819) - we acquire and drop i_flags_lock a lot and thus provide no consistency between the various flags we set/clear under it This patch fixes all that with a major revamp of the locking in the function. The new version acquires i_flags_lock early and only drops it once we need to call into inode_init_always or before calling xfs_ilock. This patch fixes a bug seen in the wild where we race modifying the reclaim tag. Signed-off-by:
-
- 17 Aug, 2009 1 commit
-
-
Christoph Hellwig authored
The locking in xfs_iget_cache_hit currently has numerous problems: - we clear the reclaim tag without i_flags_lock which protects modifications to it - we call inode_init_always which can sleep with pag_ici_lock held (this is oss.sgi.com BZ #819) - we acquire and drop i_flags_lock a lot and thus provide no consistency between the various flags we set/clear under it This patch fixes all that with a major revamp of the locking in the function. The new version acquires i_flags_lock early and only drops it once we need to call into inode_init_always or before calling xfs_ilock. This patch fixes a bug seen in the wild where we race modifying the reclaim tag. Signed-off-by:
-
- 16 Aug, 2009 2 commits
-
-
Randy Dunlap authored
drivers/md/dm-log-userspace-transfer.c:110: warning: format '%lu' expects type 'long unsigned int', but argument 4 has type 'size_t' Previously posted and acked, but apparently lost. http://lkml.indiana.edu/hypermail/linux/kernel/0906.2/02074.htmlSigned-off-by:
Randy Dunlap <randy.dunlap@oracle.com> Cc: dm-devel@redhat.com Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org>
-
Guillaume Knispel authored
The triggered field of struct poll_wqueues introduced in commit 5f820f64 ("poll: allow f_op->poll to sleep"). It was first set to 1 in pollwake() (now __pollwake() ), tested and later set to 0 in poll_schedule_timeout(), but not initialized before. As a result when the process needs to sleep, triggered was likely to be non-zero even if pollwake() is not called before the first poll_schedule_timeout(), meaning schedule_hrtimeout_range() would not be called and an extra loop calling all ->poll() would be done. This patch initialize triggered to 0 in poll_initwait() so the ->poll() are not called twice before the process goes to sleep when it needs to. Signed-off-by:
Guillaume Knispel <gknispel@proformatique.com> Acked-by:
Thomas Gleixner <tglx@linutronix.de> Acked-by:
Tejun Heo <tj@kernel.org> Cc: stable@kernel.org Signed-off-by:
-
- 14 Aug, 2009 3 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixesLinus Torvalds authored
* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes: GFS2: Fix permissions on "recover" file
-
git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6Linus Torvalds authored
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6: (38 commits) V4L/DVB (12441): siano: read buffer overflow V4L/DVB (12440): Use kzalloc for frontend states to have struct dvb_frontend properly V4L/DVB (12438): Read buffer overflow V4L/DVB (12437): dvb: siano uses/depends on INPUT V4L/DVB (12436): stk-webcam: read buffer overflow V4L/DVB (12432): em28xx: fix regression in Empire DualTV digital tuning V4L/DVB (12429): v4l2-ioctl: fix G_STD and G_PARM default handlers V4L/DVB (12428): hdpvr: add missing initialization of current_norm V4L/DVB (12424): soc-camera: fix recursive locking in .buf_queue() V4L/DVB (12422): media/zr364xx: fix build errors V4L/DVB (12405): em28xx-cards: move register 0x13 setting to the proper place V4L/DVB (12411): em28xx: Fix artifacts with Silvercrest webcam V4L/DVB (12410): em28xx: Move the non-board dependent part to be outside em28xx_pre_card_setup() V4L/DVB (12407): em28xx: Adjust Silvercrest xtal frequency V4L/DVB (12406): em28xx: fix: don't do image interlacing on webcams V4L/DVB (12403): em28xx: properly reports some em2710 chips V4L/DVB (12402): em28xx: fix: some em2710 chips use a different vendor ID V4L/DVB (12401): m9v011: add vflip/hflip controls to control mirror/upside down V4L/DVB (12400): em28xx: Allow changing fps on webcams V4L/DVB (12399): mt9v011: Add support for controlling frame rates ...
-
Steven Whitehouse authored
Although this file is only ever written and not read by userspace, it seems that the utils are opening this file O_RDWR, so we need to allow that. Also fixes the whitespace which seemed to be broken. Signed-off-by:Steven Whitehouse <swhiteho@redhat.com> Cc: David Teigland <teigland@redhat.com>
-
- 13 Aug, 2009 19 commits
-
-
Roel Kluin authored
With mode DEVICE_MODE_RAW_TUNER a read occurs past the end of smscore_fw_lkup[]. Subsequently an attempt is made to load the firmware from the resulting filename. Signed-off-by:
Roel Kluin <roel.kluin@gmail.com> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Douglas Schilling Landgraf <dougsland@redhat.com> Signed-off-by:
Mauro Carvalho Chehab <mchehab@redhat.com>
-
Matthias Schwarzott authored
This patch changes most frontend drivers to allocate their state structure via kzalloc and not kmalloc. This is done to properly initialize the embedded "struct dvb_frontend frontend" field, that they all have. The visible effect of this struct being uninitalized is, that the member "id" that is used to set the name of kernel thread is totally random. Some board drivers (for example cx88-dvb) set this "id" via videobuf_dvb_alloc_frontend but most do not. So I at least get random id values for saa7134, flexcop and ttpci based cards. It looks like this in dmesg: DVB: registering adapter 1 frontend -10551321 (ST STV0299 DVB-S) The related kernel thread then also gets a strange name like "kdvb-ad-1-fe--1". Cc: Michael Krufky <mkrufky@linuxtv.org> Cc: Steven Toth <stoth@linuxtv.org> Cc: Timothy Lee <timothy.lee@siriushk.com> Cc: Igor M. Liplianin <liplianin@me.by> Signed-off-by:
Matthias Schwarzott <zzam@gentoo.org> Acked-by:
Andreas Oberritter <obi@linuxtv.org> Signed-off-by:
-
Roel Kluin authored
parport[n] is checked before n < MAX_CAMS Signed-off-by:
-
Randy Dunlap authored
siano uses input_*() functions so it should depend on INPUT to prevent build errors: ERROR: "input_event" [drivers/media/dvb/siano/sms1xxx.ko] undefined! ERROR: "input_register_device" [drivers/media/dvb/siano/sms1xxx.ko] undefined! ERROR: "input_free_device" [drivers/media/dvb/siano/sms1xxx.ko] undefined! ERROR: "input_unregister_device" [drivers/media/dvb/siano/sms1xxx.ko] undefined! ERROR: "input_allocate_device" [drivers/media/dvb/siano/sms1xxx.ko] undefined! Cc: Michael Krufky <mkrufky@linuxtv.org> Cc: Uri Shkolnik <uris@siano-ms.com> Signed-off-by:
-
Roel Kluin authored
It tested the value of stk_sizes[i].m before checking whether i was in range. Cc: Hans Verkuil <hverkuil@xs4all.nl> Cc: Trent Piepho <xyzzy@speakeasy.org> Signed-off-by:
-
Devin Heitmueller authored
Restore support for digital tuning caused by regression during introduction of disable_i2c_gate parameter to zl10353 driver. Thanks to user "Xwang" for reporting the problem and testing the fix Cc: Xwang <xwang1976@email.it> Signed-off-by:
Devin Heitmueller <dheitmueller@kernellabs.com> Signed-off-by:
-
Hans Verkuil authored
The v4l core supplies default handlers for G_STD and G_PARM. However, both default handlers are buggy. This patch fixes the following: 1) If no g_std is supplied and current_norm == 0, then this driver does not support TV video standards (e.g. a radio or webcam driver). Return -EINVAL. This ensures that there is no bogus VIDIOC_G_STD support for such drivers. 2) The default VIDIOC_G_PARM handler used current_norm instead of first checking if the driver supported g_std and calling that to get the norm. It also didn't check if current_norm was 0, since in that case the driver does not support TV standards (or no standard was set at all) and the default handler should return -EINVAL. Note that I am very unhappy with these default handlers: I think they basically behave like some very strange and unexpected side-effect. Signed-off-by:
Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by:
-
Hans Verkuil authored
Drivers should either set current_norm or supply a g_std callback. The hdpvr driver does neither. Since it initializes to a 60 Hz format I've initialized the current_norm to NTSC | PAL_M | PAL_60 which is the 60 Hz subset of tvnorms. Cc: Janne Grunau <j@jannau.net> Signed-off-by:
-
Guennadi Liakhovetski authored
The .buf_queue() V4L2 driver method is called under spinlock_irqsave(q->irqlock,...), don't take the lock again inside the function. Reported-by:
Antonio Ospite <ospite@studenti.unina.it> Signed-off-by:
Guennadi Liakhovetski <g.liakhovetski@gmx.de> Signed-off-by:
-
Randy Dunlap authored
Fix build errors in zr364xx by adding selects: zr364xx.c:(.text+0x195ed7): undefined reference to `videobuf_streamon' zr364xx.c:(.text+0x196030): undefined reference to `videobuf_dqbuf' zr364xx.c:(.text+0x1960c4): undefined reference to `videobuf_qbuf' zr364xx.c:(.text+0x196123): undefined reference to `videobuf_querybuf' zr364xx.c:(.text+0x196182): undefined reference to `videobuf_reqbufs' zr364xx.c:(.text+0x196224): undefined reference to `videobuf_queue_is_busy' zr364xx.c:(.text+0x196390): undefined reference to `videobuf_vmalloc_free' zr364xx.c:(.text+0x196571): undefined reference to `videobuf_iolock' zr364xx.c:(.text+0x196678): undefined reference to `videobuf_mmap_mapper' zr364xx.c:(.text+0x196760): undefined reference to `videobuf_poll_stream' zr364xx.c:(.text+0x19689a): undefined reference to `videobuf_read_one' zr364xx.c:(.text+0x1969ec): undefined reference to `videobuf_mmap_free' zr364xx.c:(.text+0x197862): undefined reference to `videobuf_queue_vmalloc_init' zr364xx.c:(.text+0x197a28): undefined reference to `videobuf_streamoff' zr364xx.c:(.text+0x198203): undefined reference to `videobuf_to_vmalloc' zr364xx.c:(.text+0x198603): undefined reference to `videobuf_streamoff' drivers/built-in.o: In function `free_buffer': zr364xx.c:(.text+0x19930c): undefined reference to `videobuf_vmalloc_free' drivers/built-in.o: In function `zr364xx_open': zr364xx.c:(.text+0x19a7de): undefined reference to `videobuf_queue_vmalloc_init' drivers/built-in.o: In function `read_pipe_completion': zr364xx.c:(.text+0x19b17f): undefined reference to `videobuf_to_vmalloc' Signed-off-by:
-
Mauro Carvalho Chehab authored
Register 0x13 seems to be a sort of image control, maybe gamma, white level or black level. Lower values produce better images, while higher values increases the contrast and shifts colors to green. 0xff produces a black image. This register is not Silvercrest-specific, so its code should be moved to a better place. If this register is left alone, a random value can be found at the register, producing weird results. While here, let's remove register 0x0d, as it had no noticed effect at the image. Signed-off-by: -
Mauro Carvalho Chehab authored
Silvercrest mt9v011 sensor produces a 640x480 image. However, previously, the code were getting only half of the lines and merging two consecutive frames to "produce" a 640x480 image. With the addition of progressive mode, now em28xx is working with a full image. However, when the number of lines is bigger than 240, the beginning of some odd lines are filled with blank. After lots of testing, and physically checking the device for a Xtal, it was noticed experimentally that mt9v011 is using em28xx XCLK as its clock. Due to that, changing XCLK value changes the maximum speed of the stream. At the tests, it were possible to produce up to 32 fps, using a 30 MHz XCLK. However, at that rate, the artifacts happen even at 320x240. Lower values of XCLK produces artifacts only at 640x480. At some values of xclk (for example XCLKK = 6 MHz, 640x480), it is possible to see an invalid sucession of artifacts with this pattern: .xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ..xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ...xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ....xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx .xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ..xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ...xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ....xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx (where the dots represent the blanked pixels) So, it seems that a waveform in the format of a ramp is interferring at the image. The cause of this interference is currently unknown. Some possibilities are: - electrical interference (maybe this device is broken?); - some issue at mt9v011 programming; - some bug at em28xx chip. So, for now, let's be conservative and use a value of XCLK that we know for sure that it won't cause artifacts. As I'm waiting for more of such devices with different em28xx chipset revisions, I'll have the opportunity to double check the issue with other pieces of hardware. Later patches can vary XCLK depending on the vertical resolutions, if a proper fix is not discovered. Signed-off-by: -
Mauro Carvalho Chehab authored
em28xx_pre_card_setup() is meant to contain board-specific initialization. Also, as autodetection sometimes occur only after having i2c bus enabled, this function may need to be called later. Moving those setups to happen outside the function avoids calling it twice without need and without duplicating output lines at dmesg. Signed-off-by: -
Mauro Carvalho Chehab authored
We don't know the xtal frequency of Silvercrest, but we need to have some value in order to allow controlling the frame rate frequency. The value is probably still wrong, since the manufacturer announces this device as being capable of 30fps, but the maximum we can get is 13.5 fps. Signed-off-by: -
Mauro Carvalho Chehab authored
Due to historical reasons, em28xx driver gets two consecutive frames and fold them into an unique framing, doing interlacing. While this works fine for TV images, this produces two bad effects with webcams: 1) webcam images are progressive. Merging two consecutive images produce interlacing artifacts on the image; 2) since the driver needs to get two frames, it reduces the maximum frame rate by two. Signed-off-by: -
Mauro Carvalho Chehab authored
As reported by hermann pitton <hermann-pitton@arcor.de>, some devices has a different chip id for em2710 (likely the older ones): em28xx: New device @ 480 Mbps (eb1a:2710, interface 0, class 0) em28xx #0: Identified as EM2710/EM2750/EM2751 webcam grabber (card=22) em28xx #0: em28xx chip ID = 17 Signed-off-by: -
Mauro Carvalho Chehab authored
Thanks to hermann pitton <hermann-pitton@arcor.de> for pointing this new variation. Tested-by:
hermann pitton <hermann-pitton@arcor.de> Signed-off-by:
-
Mauro Carvalho Chehab authored
Signed-off-by: -
Mauro Carvalho Chehab authored
em28xx doesn't have temporal scaling. However, on webcams, sensors are capable of changing the output rate. So, VIDIOC_[G|S]_PARM ioctls should be passed to the sensor for it to properly set frame rate. Signed-off-by:
-
