fbdev: defio and Metronomefb v3
Subsystems (2)
PATCH 1/1 fbdev: defio and Metronomefb v3
Hi Tony, Geert, fbdev,
This patch implements support for the E-Ink Metronome controller. It provides an mmapable interface to the controller using defio support.
I welcome your feedback. Please let me know if it looks okay to apply.
Thanks,
jaya
---
Documentation/fb/cmap_xfbdev.txt | 53 ++ Documentation/fb/metronomefb.txt | 38 + drivers/video/Kconfig | 23 drivers/video/Makefile | 1 drivers/video/fb_defio.c | 22 drivers/video/metronomefb.c | 975 +++++++++++++++++++++++++++++++++++++++ 6 files changed, 1110 insertions(+), 2 deletions(-)
Documentation/fb/cmap_xfbdev.txt new file +53 -0
@@ -0,0 +1,53 @@ +Understanding fbdev's cmap +-------------------------- + +These notes explain how X's dix layer uses fbdev's cmap structures. + +*. example of relevant structures in fbdev as used for a 3-bit grayscale cmap +struct fb_var_screeninfo { + .bits_per_pixel = 8, + .grayscale = 1, + .red = { 4, 3, 0 }, + .green = { 0, 0, 0 }, + .blue = { 0, 0, 0 }, +} +struct fb_fix_screeninfo { + .visual = FB_VISUAL_STATIC_PSEUDOCOLOR, +} +for (i = 0; i < 8; i++) + info->cmap.red[i] = (((2*i)+1)*(0xFFFF))/16; +memcpy(info->cmap.green, info->cmap.red, sizeof(u16)*8); +memcpy(info->cmap.blue, info->cmap.red, sizeof(u16)*8); + +*. X11 apps do something like the following when trying to use grayscale. +for (i=0; i < 8; i++) { + char colorspec[64]; + memset(colorspec,0,64); + sprintf(colorspec, "rgb:%x/%x/%x", i*36,i*36,i*36); + if (!XParseColor(outputDisplay, testColormap, colorspec, &wantedColor)) + printf("Can't get color %s\n",colorspec); + XAllocColor(outputDisplay, testColormap, &wantedColor); + grays[i] = wantedColor; +} +There's also named equivalents like gray1..x provided you have an rgb.txt. + +Somewhere in X's callchain, this results in a call to X code that handles the +colormap. For example, Xfbdev hits the following: + +xc-011010/programs/Xserver/dix/colormap.c: + +FindBestPixel(pentFirst, size, prgb, channel) + +dr = (long) pent->co.local.red - prgb->red; +dg = (long) pent->co.local.green - prgb->green; +db = (long) pent->co.local.blue - prgb->blue; +sq = dr * dr; +UnsignedToBigNum (sq, &sum); +BigNumAdd (&sum, &temp, &sum); + +co.local.red are entries that were brought in through FBIOGETCMAP which come +directly from the info->cmap.red that was listed above. The prgb is the rgb +that the app wants to match to. The above code is doing what looks like a least +squares matching function. That's why the cmap entries can't be set to the left +hand side boundaries of a color range. +
Documentation/fb/metronomefb.txt new file +38 -0
@@ -0,0 +1,38 @@ + Metronomefb + ----------- +Maintained by Jaya Kumar <jayakumar.lkml.gmail.com> +Last revised: Nov 20, 2007 + +Metronomefb is a driver for the Metronome display controller. The controller +is from E-Ink Corporation. It is intended to be used to drive the E-Ink +Vizplex display media. E-Ink hosts some details of this controller and the +display media here http://www.e-ink.com/products/matrix/metronome.html . + +Metronome is interfaced to the host CPU through the AMLCD interface. The +host CPU generates the control information and the image in a framebuffer +which is then delivered to the AMLCD interface by a host specific method. +Currently, that's implemented for the PXA's LCDC controller. The display and +error status are each pulled through individual GPIOs. + +Metronomefb was written for the PXA255/gumstix/lyre combination and +therefore currently has board set specific code in it. If other boards based on +other architectures are available, then the host specific code can be separated +and abstracted out. + +Metronomefb requires waveform information which is delivered via the AMLCD +interface to the metronome controller. The waveform information is expected to +be delivered from userspace via the firmware class interface. The waveform file +can be compressed as long as your udev or hotplug script is aware of the need +to uncompress it before delivering it. metronomefb will ask for waveform.wbf +which would typically go into /lib/firmware/waveform.wbf depending on your +udev/hotplug setup. I have only tested with a single waveform file which was +originally labeled 23P01201_60_WT0107_MTC. I do not know what it stands for. +Caution should be exercised when manipulating the waveform as there may be +a possibility that it could have some permanent effects on the display media. +I neither have access to nor know exactly what the waveform does in terms of +the physical media. + +Metronomefb uses the deferred IO interface so that it can provide a memory +mappable frame buffer. It has been tested with tinyx (Xfbdev). It is known +to work at this time with xeyes, xclock, xloadimage, xpdf. +
drivers/video/Kconfig +14 -0
@@ -1880,6 +1889,20 @@ config FB_XILINX framebuffer. ML300 carries a 640*480 LCD display on the board, ML403 uses a standard DB15 VGA connector. +config FB_METRONOME + tristate "Metronome display controller support" + depends on FB && ARCH_PXA && MMU + select FB_SYS_FILLRECT + select FB_SYS_COPYAREA + select FB_SYS_IMAGEBLIT + select FB_SYS_FOPS + select FB_DEFERRED_IO + help + This enables support for the Metronome display controller. Tested + with an E-Ink 800x600 display and Gumstix Connex through an AMLCD + interface. Please read <file:Documentation/fb/metronomefb.txt> + for more information. + config FB_VIRTUAL tristate "Virtual Frame Buffer support (ONLY FOR TESTING!)" depends on FB
drivers/video/Makefile +1 -0
@@ -103,6 +103,7 @@ obj-$(CONFIG_FB_PMAG_AA) += pmag-aa-fb.o obj-$(CONFIG_FB_PMAG_BA) += pmag-ba-fb.o obj-$(CONFIG_FB_PMAGB_B) += pmagb-b-fb.o obj-$(CONFIG_FB_MAXINE) += maxinefb.o +obj-$(CONFIG_FB_METRONOME) += metronomefb.o obj-$(CONFIG_FB_S1D13XXX) += s1d13xxxfb.o obj-$(CONFIG_FB_IMX) += imxfb.o obj-$(CONFIG_FB_S3C2410) += s3c2410fb.o
drivers/video/fb_defio.c +20 -2
@@ -4,7 +4,7 @@* Copyright (C) 2006 Jaya Kumar * * This file is subject to the terms and conditions of the GNU General Public -* License. See the file COPYING in the main directory of this archive +* License. See the file COPYING in the main directory of this archive * for more details. */
@@ -31,7 +31,7 @@ static int fb_deferred_io_fault(struct vm_area_struct *vma,unsigned long offset;struct page *page;struct fb_info *info = vma->vm_private_data; -/* info->screen_base is in System RAM */ +/* info->screen_base is virtual memory */ void *screen_base = (void __force *) info->screen_base; offset = vmf->pgoff <<PAGE_SHIFT ;
@@ -43,6 +43,15 @@ static int fb_deferred_io_fault(struct vm_area_struct *vma,return VM_FAULT_SIGBUS ;get_page (page); + +if (vma->vm_file) + page->mapping = vma->vm_file->f_mapping; +else +printk (KERN_ERR "no mapping available\n" ); + +BUG_ON (!page->mapping); + page->index = vmf->pgoff; + vmf->page = page;return 0 ; }
@@ -138,11 +147,20 @@ EXPORT_SYMBOL_GPL(fb_deferred_io_init);void fb_deferred_io_cleanup (struct fb_info *info) { +void *screen_base = (void __force *) info->screen_base;struct fb_deferred_io *fbdefio = info->fbdefio; +struct page *page; +int i;BUG_ON (!fbdefio);cancel_delayed_work (&info->deferred_work);flush_scheduled_work (); + +/* clear out the mapping that we setup */ +for (i =0 ; i < info->fix.smem_len; i +=PAGE_SIZE ) { + page =vmalloc_to_page (screen_base + i); + page->mapping =NULL ; + } }EXPORT_SYMBOL_GPL (fb_deferred_io_cleanup);
drivers/video/metronomefb.c new file +975 -0
975 lines not on this page · raw message
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 4 lines · Jaya Kumar
This patch implements support for the E-Ink Metronome controller. It provides an mmapable interface to the controller using defio support.
I welcome your feedback. Please let me know if it looks okay to apply.
To me it looks OK.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 8 lines · Geert Uytterhoeven
Quoted 4 lines · Jaya Kumar
This patch implements support for the E-Ink Metronome controller. It provides an mmapable interface to the controller using defio support.
I welcome your feedback. Please let me know if it looks okay to apply.
To me it looks OK.
Thanks Geert.
I have a follow-up question on the next steps of getting this merged. I saw Stephen Rothwell's announcement of a new tree called linux-next. [ http://lwn.net/Articles/268881/ ] He also said:
"
otoh, a lot of the -mm queue is _not_ cross-subsystem: fbdev, uml, various new drivers and features, etc. So I'll need to split -mm into multiple queues and feed many of them into linux-next.
"
I have not understood this. Does this mean that this type of patch,
ie: a new fbdev driver, should aim towards merging into akpm's -mm tree? or that there needs to be a fbdev subsystem git tree that Stephen can pull from into linux-next?
Thanks,
jaya
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 13 lines · On Wed, 20 Feb 2008 08:18:35 -0500 "Jaya Kumar" <[email protected]>…
Quoted 8 lines · Geert Uytterhoeven
Quoted 4 lines · Jaya Kumar
This patch implements support for the E-Ink Metronome controller. It provides an mmapable interface to the controller using defio support.
I welcome your feedback. Please let me know if it looks okay to apply.
To me it looks OK.
Thanks Geert.
Quoted 13 lines
I have a follow-up question on the next steps of getting this merged. I saw Stephen Rothwell's announcement of a new tree called linux-next. [ http://lwn.net/Articles/268881/ ] He also said:
"
otoh, a lot of the -mm queue is _not_ cross-subsystem: fbdev, uml, various new drivers and features, etc. So I'll need to split -mm into multiple queues and feed many of them into linux-next.
"
I have not understood this. Does this mean that this type of patch,
ie: a new fbdev driver, should aim towards merging into akpm's -mm tree? or that there needs to be a fbdev subsystem git tree that Stephen can pull from into linux-next?
It means that at some stage I'd like linux-next to incorporate some patches from the -mm queue, that's all.
I'll continue to collect fbdev patches.
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 6 lines · Andrew Morton
It means that at some stage I'd like linux-next to incorporate some patches from the -mm queue, that's all.
I'll continue to collect fbdev patches.
Ok. Cool. Does this metronomefb patch look viable for merging into -mm?
Thanks,
jaya
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 11 lines · On Wed, 20 Feb 2008 08:46:17 -0500 "Jaya Kumar" <[email protected]>…
Quoted 6 lines · Andrew Morton
It means that at some stage I'd like linux-next to incorporate some patches from the -mm queue, that's all.
I'll continue to collect fbdev patches.
Ok. Cool. Does this metronomefb patch look viable for merging into -mm?
I haven't had a chance to sit down and look through it yet. I shall do so.
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 8 lines · On Mon, 18 Feb 2008 15:03:08 +0100 (CET) Geert Uytterhoeven…
Quoted 4 lines · Jaya Kumar
This patch implements support for the E-Ink Metronome controller. It provides an mmapable interface to the controller using defio support.
I welcome your feedback. Please let me know if it looks okay to apply.
To me it looks OK.
It doesn't to me - I think I found several issues.
Please, silence is preferable to a quick scan and a poor ack.
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 24 lines · On Mon, 18 Feb 2008 08:41:26 -0500 Jaya Kumar <[email protected]> wrote:
Hi Tony, Geert, fbdev,
This patch implements support for the E-Ink Metronome controller. It provides an mmapable interface to the controller using defio support.
I welcome your feedback. Please let me know if it looks okay to apply.
...
drivers/video/fb_defio.c +3
-1@@ -31,7 +31,7 @@ static int fb_deferred_io_fault(struct vm_area_struct *vma,unsigned long offset;struct page *page;struct fb_info *info = vma->vm_private_data; -/* info->screen_base is in System RAM */ +/* info->screen_base is virtual memory */ void *screen_base = (void __force *) info->screen_base; offset = vmf->pgoff <<PAGE_SHIFT ;@@ -43,6 +43,15 @@ static int fb_deferred_io_fault(struct vm_area_struct *vma,return VM_FAULT_SIGBUS ;get_page (page); + +if (vma->vm_file)
drivers/video/fb_defio.c static int fb_deferred_io_fault(struct vm_area_struct *vma, · line 57
Quoted 10 lines
drivers/video/fb_defio.c +7
-0+ page->mapping = vma->vm_file->f_mapping; + else + printk(KERN_ERR "no mapping available\n"); + +BUG_ON (!page->mapping); + page->index = vmf->pgoff; + vmf->page = page;return 0 ; }
What locking prevents `page' from being stripped off its mapping here? y say munmap or truncate (if it's supported here, which it presumably isn't).
Quoted 9 lines
drivers/video/fb_defio.c +3
-0@@ -138,11 +147,20 @@ EXPORT_SYMBOL_GPL(fb_deferred_io_init);void fb_deferred_io_cleanup (struct fb_info *info) { +void *screen_base = (void __force *) info->screen_base;struct fb_deferred_io *fbdefio = info->fbdefio; +struct page *page; +int i;
drivers/video/fb_defio.c EXPORT_SYMBOL_GPL(fb_deferred_io_init); · line 164
Quoted 10 lines
drivers/video/fb_defio.c +6
-0BUG_ON (!fbdefio);cancel_delayed_work (&info->deferred_work);flush_scheduled_work (); + +/* clear out the mapping that we setup */ +for (i =0 ; i < info->fix.smem_len; i +=PAGE_SIZE ) { + page =vmalloc_to_page (screen_base + i); + page->mapping =NULL ; + } }
ie: a race with this function?
drivers/video/metronomefb.c line 165
Quoted 1 line
drivers/video/metronomefb.c new file +1
-0+int load_waveform (u8 *mem,u8 *metromem,int m,int t,u8 *frame_count)
This didn't need to be a global symbol. It would be a poorly chosen idenfitier if it was global.
Quoted 59 lines
drivers/video/metronomefb.c new file +59
-0+{ +int tta; +int wmta; +int trn =0 ; +int i; +unsigned char v; +u8 cksum; +int cksum_idx; +int wfm_idx, owfm_idx; +int mem_idx =0 ; +struct waveform_hdr *wfm_hdr; + + wfm_hdr = (struct waveform_hdr *) mem; + +if (wfm_hdr->fvsn !=1 ) { +printk (KERN_ERR "Error: bad fvsn %x\n" , wfm_hdr->fvsn); +return -EINVAL ; + } +if (wfm_hdr->luts !=0 ) { +printk (KERN_ERR "Error: bad luts %x\n" , wfm_hdr->luts); +return -EINVAL ; + } + cksum =calc_cksum (32 ,47 , mem); +if (cksum != wfm_hdr->wfm_cs) { +printk (KERN_ERR "Error: bad cksum %x != %x\n" , cksum, + wfm_hdr->wfm_cs); +return -EINVAL ; + } + wfm_hdr->mc +=1 ; + wfm_hdr->trc +=1 ; +for (i =0 ; i <5 ; i++) { +if (*(wfm_hdr->stuff2a + i) !=0 ) { +printk (KERN_ERR "Error: unexpected value in padding\n" ); +return -EINVAL ; + } + } + +/* calculating trn. trn is something used to index into +the waveform. presumably selecting the right one for the +desired temperature. it works out the offset of the first +v that exceeds the specified temperature */ +for (i =sizeof (*wfm_hdr); i <=sizeof (*wfm_hdr) + wfm_hdr->trc; i++) { +if (mem[i] > t) { + trn = i -sizeof (*wfm_hdr) -1 ; +break ; + } + } + +/* check temperature range table checksum */ + cksum_idx =sizeof (*wfm_hdr) + wfm_hdr->trc +1 ; + cksum =calc_cksum (sizeof (*wfm_hdr), cksum_idx, mem); +if (cksum != mem[cksum_idx]) { +printk (KERN_ERR "Error: bad temperature range table cksum" +" %x != %x\n" , cksum, mem[cksum_idx]); +return -EINVAL ; + } + +/* check waveform mode table address checksum */ + wmta = *((int *) wfm_hdr->wmta) &0x00FFFFFF ;
drivers/video/metronomefb.c line 234
Quoted 10 lines
drivers/video/metronomefb.c new file +10
-0+ cksum_idx = wmta + m*4 +3 ; + cksum =calc_cksum (cksum_idx -3 , cksum_idx, mem); +if (cksum != mem[cksum_idx]) { +printk (KERN_ERR "Error: bad mode table address cksum" +" %x != %x\n" , cksum, mem[cksum_idx]); +return -EINVAL ; + } + +/* check waveform temperature table address checksum */ + tta = *((int *) (mem + wmta + m*4 )) &0x00FFFFFF ;
Does this code have any unaligned-access issues on non-x86?
Quoted 48 lines
drivers/video/metronomefb.c new file +45
-0+ cksum_idx = tta + trn*4 +3 ; + cksum =calc_cksum (cksum_idx -3 , cksum_idx, mem); +if (cksum != mem[cksum_idx]) { +printk (KERN_ERR "Error: bad temperature table address cksum" +" %x != %x\n" , cksum, mem[cksum_idx]); +return -EINVAL ; + } + +/* here we do the real work of putting the waveform into the +metromem buffer. this does runlength decoding of the waveform */ + owfm_idx = wfm_idx = *((int *) (mem + tta + trn*4 )) &0x00FFFFFF ; +while (1 ) { +unsigned char rl; + v = mem[wfm_idx++]; +if (v == wfm_hdr->swtb) { +while ((v = mem[wfm_idx++]) != wfm_hdr->swtb) + metromem[mem_idx++] = v; + +continue ; + } + +if (v == wfm_hdr->endb) +break ; + + rl = mem[wfm_idx++]; +for (i =0 ; i <= rl; i++) + metromem[mem_idx++] = v; + } + + cksum_idx = wfm_idx; + cksum =calc_cksum (owfm_idx, cksum_idx, mem); +if (cksum != mem[cksum_idx]) { +printk (KERN_ERR "Error: bad waveform data cksum" +" %x != %x\n" , cksum, mem[cksum_idx]); +return -EINVAL ; + } + *frame_count = (mem_idx/64 ); + +return 0 ; +} +/* this technique copied from pxafb */ +static void metronome_disable_lcd_controller (struct metronomefb_par *par) +{ +DECLARE_WAITQUEUE (wait, current); +
drivers/video/metronomefb.c line 356
Quoted 10 lines
drivers/video/metronomefb.c new file +10
-0+set_current_state (TASK_UNINTERRUPTIBLE ); +add_wait_queue (&par->waitq, &wait); + +LCSR =0xffffffff ;/* Clear LCD Status Register */ +LCCR0 &= ~LCCR0_LDM ;/* Enable LCD Disable Done Interrupt */ +LCCR0 |=LCCR0_DIS ;/* Disable LCD Controller */ + +schedule_timeout (200 *HZ /1000 ); +remove_wait_queue (&par->waitq, &wait); +}
I assume that if the schedule_timeout() actually times out, we have an error condition? Should it be reported?
Quoted 35 lines
drivers/video/metronomefb.c new file +32
-0+static void metronome_enable_lcd_controller (struct metronomefb_par *par) +{ +LCSR =0xffffffff ; +FDADR0 = par->metromem_desc_dma; +LCCR0 |=LCCR0_ENB ; +} +static int metronome_display_cmd (struct metronomefb_par *par) +{ +int i; +u16 cs; +u16 opcode; +static u8 borderval; +u8 *ptr; + +/* setup display command +we can't immediately set the opcode since the controller +will try parse the command before we've set it all up +so we just set cs here and set the opcode at the end */ + + ptr = par->metromem; + +if (par->metromem_cmd->opcode ==0xCC40 ) + opcode = cs =0xCC41 ; +else + opcode = cs =0xCC40 ; + +/* set the args ( 2 bytes ) for display */ + i =0 ; + par->metromem_cmd->args[i] =1 <<3 /* border update */ + | ((borderval++ %4 ) &0x0F ) <<4 + | (par->frame_count -1 ) <<8 ; + cs += par->metromem_cmd->args[i++];
drivers/video/metronomefb.c line 477
Quoted 10 lines
drivers/video/metronomefb.c new file +10
-0+ +/* the rest are 0 */ +memset ((u8 *) (par->metromem_cmd->args + i),0 , (32 -i)*2 ); + + par->metromem_cmd->csum = cs; + par->metromem_cmd->opcode = opcode;/* display cmd */ + + i =wait_event_interruptible_timeout (par->waitq, (GPLR1 &0x01 ),HZ ); +return i; +}
wait_event_interruptible_timeout() will return immediately if the calling task has a signal pending. Will this ause the driver to malfunction?
Quoted 19 lines
drivers/video/metronomefb.c new file +19
-0+static int __devinit metronome_powerup_cmd (struct metronomefb_par *par) +{ +int i; +u16 cs; + +/* setup power up command */ + par->metromem_cmd->opcode =0x1234 ;/* pwr up pseudo cmd */ + cs = par->metromem_cmd->opcode; + +/* set pwr1,2,3 to 1024 */ +for (i =0 ; i <3 ; i++) { + par->metromem_cmd->args[i] =1024 ; + cs += par->metromem_cmd->args[i]; + } + +/* the rest are 0 */ +memset ((u8 *) (par->metromem_cmd->args + i),0 , (32 -i)*2 ); + + par->metromem_cmd->csum = cs;
drivers/video/metronomefb.c line 507
Quoted 10 lines
drivers/video/metronomefb.c new file +10
-0+ +msleep (1 ); +metronome_set_gpio_output (RST_GPIO_PIN ,1 ); + +msleep (1 ); +metronome_set_gpio_output (STDBY_GPIO_PIN ,1 ); + + i =wait_event_interruptible_timeout (par->waitq, (GPLR1 &0x01 ),HZ ); +return i; +}
Ditto
Quoted 35 lines
drivers/video/metronomefb.c new file +35
-0+static int __devinit metronome_config_cmd (struct metronomefb_par *par) +{ +int i; +u16 cs; + +/* setup config command +we can't immediately set the opcode since the controller +will try parse the command before we've set it all up +so we just set cs here and set the opcode at the end */ + + cs =0xCC10 ; + +/* set the 12 args ( 8 bytes ) for config. see spec for meanings */ + i =0 ; + par->metromem_cmd->args[i] =15 /* sdlew */ + |2 <<8 /* sdosz */ + |0 <<11 /* sdor */ + |0 <<12 /* sdces */ + |0 <<15 ;/* sdcer */ + cs += par->metromem_cmd->args[i++]; + + par->metromem_cmd->args[i] =42 /* gdspl */ + |1 <<8 /* gdr1 */ + |1 <<9 /* sdshr */ + |0 <<15 ;/* gdspp */ + cs += par->metromem_cmd->args[i++]; + + par->metromem_cmd->args[i] =18 /* gdspw */ + |0 <<15 ;/* dispc */ + cs += par->metromem_cmd->args[i++]; + + par->metromem_cmd->args[i] =599 /* vdlc */ + |0 <<11 /* dsi */ + |0 <<12 ;/* dsic */ + cs += par->metromem_cmd->args[i++];
drivers/video/metronomefb.c line 553
Quoted 10 lines
drivers/video/metronomefb.c new file +10
-0+ +/* the rest are 0 */ +memset ((u8 *) (par->metromem_cmd->args + i),0 , (32 -i)*2 ); + + par->metromem_cmd->csum = cs; + par->metromem_cmd->opcode =0xCC10 ;/* config cmd */ + + i =wait_event_interruptible_timeout (par->waitq, (GPLR1 &0x01 ),HZ ); +return i; +}
more...
Quoted 25 lines
drivers/video/metronomefb.c new file +25
-0+/* +* this is the slow path from userspace. they can seek and write to +* the fb. +*/ +static ssize_t metronomefb_write (struct fb_info *info,const char __user *buf, +size_t count,loff_t *ppos) +{ +unsigned long p; +int err = -EINVAL ; +struct metronomefb_par *par; +unsigned int xres; +unsigned int fbmemlength; + + p = *ppos; + par = info->par; + xres = info->var.xres; + fbmemlength = (xres * info->var.yres); + +if (p > fbmemlength) +return -ENOSPC ; + + err =0 ; +if ((count + p) > fbmemlength) { + count = fbmemlength - p; + err = -ENOSPC ;
drivers/video/metronomefb.c line 716
Quoted 10 lines
drivers/video/metronomefb.c new file +10
-0+ } + +if (count) { +char *base_addr; + + base_addr = (char __force *)info->screen_base; + count -=copy_from_user (base_addr + p, buf, count); + *ppos += count; + err = -EFAULT ; + }
This looks odd. We ignore the return value from copy_from_user(), which will wreck *ppos. We also always return -EFAULT.
drivers/video/metronomefb.c line 725
Quoted 8 lines
drivers/video/metronomefb.c new file +8
-0+metronomefb_dpy_update (par); + +if (count) +return count; + +return err; +} +
I don't think I'll apply this one yet.
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 14 lines · Andrew Morton
Quoted 13 lines · On Mon, 18 Feb 2008 08:41:26 -0500 Jaya Kumar <[email protected]> wrote:
drivers/video/fb_defio.c +1
-1@@ -31,7 +31,7 @@ static int fb_deferred_io_fault(struct vm_area_struct *vma,unsigned long offset;struct page *page;struct fb_info *info = vma->vm_private_data; -/* info->screen_base is in System RAM */ +/* info->screen_base is virtual memory */ void *screen_base = (void __force *) info->screen_base; offset = vmf->pgoff <<PAGE_SHIFT ;@@ -43,6 +43,15 @@ static int fb_deferred_io_fault(struct vm_area_struct *vma,return VM_FAULT_SIGBUS ;
Quoted 16 lines
drivers/video/fb_defio.c static int fb_deferred_io_fault(struct vm_area_struct *vma, · line 57
Quoted 13 lines
drivers/video/fb_defio.c +9
-0get_page (page); + +if (vma->vm_file) + page->mapping = vma->vm_file->f_mapping; +else +printk (KERN_ERR "no mapping available\n" ); + +BUG_ON (!page->mapping); + page->index = vmf->pgoff; + vmf->page = page;return 0 ; }What locking prevents `page' from being stripped off its mapping here? y say munmap or truncate (if it's supported here, which it presumably isn't).
Hi Andrew,
I feel I need some help in understanding the details of the above issue. Here's what I had thought:
I set page->mapping here so that page_mkclean can find the right vma associated with that struct page. page->mapping, and also index are used during the deferred io callback that can occur after a write access to the page. So I must check that munmap/mremap/truncate do not affect page->mapping or index. At the moment, I am looking through do_munmap's detach_vmas_to_be_unmapped(), unmap_region(), remove_vma_list() and I think these do not cause page->mapping, index to be affected.
The defio_cleanup below is called by the metronomefb driver's remove() at rmmod time. I think remove() can't be called in a race with the above callback because I believe the driver's module->ref[]count is nonzero until the app closes(). Here's the list of scenarios I am thinking about:
1- the driver loads, then app opens, mmaps, writes, sleeps. then another app rmmods the driver. the driver won't rmmod because refcount is nonzero so the prior callback from the write can still use the page 2- same but app closes and rmmod driver before the callback is scheduled. here the defio_cleanup calls flush_scheduled_work() to make sure the deferred work is completed before we then remove the mappings. so I think this ok.
To help myself see the sequence:
- driver loads. vmalloc's the memory.
- X opens (/dev/fb1) , calls mmap
- vma with mapping is created
- X writes to a user virtual address
- defio fault handler vmalloc_to_page sets up user ptes. defio fault handler sets page's mapping above and sets up deferred callback
- X calls munmap/mremap/truncate
- do_munmap modifies the app's vma and user PTEs. i think page->mapping and page->index are untouched
- defio gets callback and mkclean()s each page so mapping must also remain untouched. then passes pagelist to driver
- driver receives its callback and walks the passed pagelist using the index value. as long as the index value in page was not changed by do_munmap activity then there's no problem here.
- rmmod is called but can't proceed because refcount is nonzero
- X calls close()
- driver calls defio cleanup which cleans up the page->mapping and then driver calls vfree.
I hope I didn't form a mistaken assumption above. I appreciate your questions on this code.
Thanks,
jaya
Quoted 11 lines
Quoted 11 lines
drivers/video/fb_defio.c +3
-0@@ -138,11 +147,20 @@ EXPORT_SYMBOL_GPL(fb_deferred_io_init);void fb_deferred_io_cleanup (struct fb_info *info) { +void *screen_base = (void __force *) info->screen_base;struct fb_deferred_io *fbdefio = info->fbdefio; +struct page *page; +int i;
Quoted 14 lines
drivers/video/fb_defio.c EXPORT_SYMBOL_GPL(fb_deferred_io_init); · line 164
Quoted 10 lines
drivers/video/fb_defio.c +6
-0BUG_ON (!fbdefio);cancel_delayed_work (&info->deferred_work);flush_scheduled_work (); + +/* clear out the mapping that we setup */ +for (i =0 ; i < info->fix.smem_len; i +=PAGE_SIZE ) { + page =vmalloc_to_page (screen_base + i); + page->mapping =NULL ; + } }ie: a race with this function?
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 8 lines · Andrew Morton
Quoted 1 line · On Mon, 18 Feb 2008 08:41:26 -0500 Jaya Kumar <[email protected]> wrote:
+int
load_waveform(u8 *mem, u8 *metromem, int m, int t, u8 *frame_count)This didn't need to be a global symbol. It would be a poorly chosen idenfitier if it was global.
Agreed. Sorry, I really should have caught that as I've made this mistake before. Will fix.
Quoted 6 lines
drivers/video/metronomefb.c line 234
Quoted 3 lines
drivers/video/metronomefb.c new file +3
-0+ +/* check waveform temperature table address checksum */ + tta = *((int *) (mem + wmta + m*4 )) &0x00FFFFFF ;Does this code have any unaligned-access issues on non-x86?
Agreed. I'll put in some checks to make sure that the indices align and also range checking to make sure we do not access beyond the buffer.
Quoted 11 lines
Quoted 11 lines
drivers/video/metronomefb.c new file +10
-0+/* this technique copied from pxafb */ +static void metronome_disable_lcd_controller (struct metronomefb_par *par) +{ +DECLARE_WAITQUEUE (wait, current); + +set_current_state (TASK_UNINTERRUPTIBLE ); +add_wait_queue (&par->waitq, &wait); + +LCSR =0xffffffff ;/* Clear LCD Status Register */ +LCCR0 &= ~LCCR0_LDM ;/* Enable LCD Disable Done Interrupt */
Quoted 8 lines
drivers/video/metronomefb.c line 392
Quoted 5 lines
drivers/video/metronomefb.c new file +5
-0+LCCR0 |=LCCR0_DIS ;/* Disable LCD Controller */ + +schedule_timeout (200 *HZ /1000 ); +remove_wait_queue (&par->waitq, &wait); +}I assume that if the
schedule_timeout()actually times out, we have an error condition? Should it be reported?
Not sure yet, I'll check.
Quoted 31 lines
Quoted 31 lines
drivers/video/metronomefb.c new file +31
-0+static int metronome_display_cmd (struct metronomefb_par *par) +{ +int i; +u16 cs; +u16 opcode; +static u8 borderval; +u8 *ptr; + +/* setup display command +we can't immediately set the opcode since the controller +will try parse the command before we've set it all up +so we just set cs here and set the opcode at the end */ + + ptr = par->metromem; + +if (par->metromem_cmd->opcode ==0xCC40 ) + opcode = cs =0xCC41 ; +else + opcode = cs =0xCC40 ; + +/* set the args ( 2 bytes ) for display */ + i =0 ; + par->metromem_cmd->args[i] =1 <<3 /* border update */ + | ((borderval++ %4 ) &0x0F ) <<4 + | (par->frame_count -1 ) <<8 ; + cs += par->metromem_cmd->args[i++]; + +/* the rest are 0 */ +memset ((u8 *) (par->metromem_cmd->args + i),0 , (32 -i)*2 ); + + par->metromem_cmd->csum = cs;
Quoted 9 lines
drivers/video/metronomefb.c line 477
Quoted 5 lines
drivers/video/metronomefb.c new file +5
-0+ par->metromem_cmd->opcode = opcode;/* display cmd */ + + i =wait_event_interruptible_timeout (par->waitq, (GPLR1 &0x01 ),HZ ); +return i; +}
wait_event_interruptible_timeout()will return immediately if the calling task has a signal pending. Will this ause the driver to malfunction?
Not sure yet. Will check.
Quoted 31 lines
Quoted 31 lines
drivers/video/metronomefb.c new file +30
-0+/* +* this is the slow path from userspace. they can seek and write to +* the fb. +*/ +static ssize_t metronomefb_write (struct fb_info *info,const char __user *buf, +size_t count,loff_t *ppos) +{ +unsigned long p; +int err = -EINVAL ; +struct metronomefb_par *par; +unsigned int xres; +unsigned int fbmemlength; + + p = *ppos; + par = info->par; + xres = info->var.xres; + fbmemlength = (xres * info->var.yres); + +if (p > fbmemlength) +return -ENOSPC ; + + err =0 ; +if ((count + p) > fbmemlength) { + count = fbmemlength - p; + err = -ENOSPC ; + } + +if (count) { +char *base_addr; +
Quoted 8 lines
drivers/video/metronomefb.c line 716
Quoted 5 lines
drivers/video/metronomefb.c new file +5
-0+ base_addr = (char __force *)info->screen_base; + count -=copy_from_user (base_addr + p, buf, count); + *ppos += count; + err = -EFAULT ; + }This looks odd. We ignore the return value from
copy_from_user(), which will wreck *ppos. We also always return-EFAULT.
Oops. Will fix.
Quoted 3 lines
I don't think I'll apply this one yet.
Yes, I agree. :-) I will fix the problems. I am grateful for the feedback.
Thanks,
jaya
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
Quoted 44 lines · Jaya Kumar
Quoted 41 lines · Andrew Morton
Quoted 36 lines · On Mon, 18 Feb 2008 08:41:26 -0500 Jaya Kumar <[email protected]> wrote:
drivers/video/metronomefb.c new file +36
-0+static int metronome_display_cmd (struct metronomefb_par *par) +{ +int i; +u16 cs; +u16 opcode; +static u8 borderval; +u8 *ptr; + +/* setup display command +we can't immediately set the opcode since the controller +will try parse the command before we've set it all up +so we just set cs here and set the opcode at the end */ + + ptr = par->metromem; + +if (par->metromem_cmd->opcode ==0xCC40 ) + opcode = cs =0xCC41 ; +else + opcode = cs =0xCC40 ; + +/* set the args ( 2 bytes ) for display */ + i =0 ; + par->metromem_cmd->args[i] =1 <<3 /* border update */ + | ((borderval++ %4 ) &0x0F ) <<4 + | (par->frame_count -1 ) <<8 ; + cs += par->metromem_cmd->args[i++]; + +/* the rest are 0 */ +memset ((u8 *) (par->metromem_cmd->args + i),0 , (32 -i)*2 ); + + par->metromem_cmd->csum = cs; + par->metromem_cmd->opcode = opcode;/* display cmd */ + + i =wait_event_interruptible_timeout (par->waitq, (GPLR1 &0x01 ),HZ ); +return i; +}
Quoted 5 lines
Quoted 3 lines
wait_event_interruptible_timeout()will return immediately if the calling task has a signal pending. Will this ause the driver to malfunction?Not sure yet. Will check.
The typical path for display_cmd is through the defio callback via the mmap user. In this scenario, it is called from keventd rather than the mmap user's context. Alternatively, display_cmd is called from the write()/cfb path in the context of the user process. In either scenario, a pending signal would cause us to return. The display command would have been set telling the metronome controller to update the display but the process would not be wait-ed. The implication of this is that we could issue a new display cmd or do display work prior to completion of the current display cmd. If we write to the framebuffer during this time, the effect is that the checksum on the transferred framebuffer data will not match and thus the controller ignores the original display cmd, and will thus only process the last display cmd which is the desired result anyway. Therefore, I think this is okay.
The remaining *_cmd functions that use wait_event_interruptible_timeout are all called from init_regs via probe() so any signal results in the allocated resources being freed by the failout in probe() and the error being returned which seems okay. But after thinking about it, I'm changing these to use wait_event_timeout since it doesn't make sense to be interruptible at those junctures.
Thanks,
jaya
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/