November 2007 - Alsa-devel - mailman.alsa-project.org

[alsa-devel] I could maybe help for Creative Labs SB X-Fi integration
by Frederic Bouy 10 Nov '07

10 Nov '07

Hello everybody, I have just installed a distribution of Ubuntu (7.10 Gutsy Gibbon) on my computer. The issue is that my sound card is not working. According to your website, this is because my SB X-Fi is not yet supported by ALSA : "Card delivered to developers. Completely new architecture. Creative actively preventing support due to no datasheets being released to ALSA developers. Reverse engineering work not started due to lack of time." I would like to know if there is a way for me to help you integrating this sound card. What can I do to help? Thanks. Frédéric

2 1

[alsa-devel] Getting XRUN after writei in RUNNING state?
by Piers O'Hanlon 10 Nov '07

10 Nov '07

Hi folks, I've trying to improve the ALSA driver for the Robust Audio Tool (RAT) tool: http://frostie.cs.ucl.ac.uk/nets/mmedia/wiki/RatWiki Specifically the code I'm talking about is: http://frostie.cs.ucl.ac.uk/nets/mmedia/browser/rat/trunk/auddev_alsa.c I've added some status dumps in the code that I'm running which aren't in the repository version but basically the code is doing the same thing. However I'm seeing weird XRUNs (under runs on the playback side) - There's as couple of instances I don't understand: (The pcm device setup is shown at the bottom); 1) RAT is happily writing audio to the playback device for while then the device just seems to go in XRUN state after being in RUNNING state and doing a writei()??: [+945 auddev_alsa.c] Time since last write: 42734us, Trying to write 640 frames state : RUNNING trigger_time: 1194622844.333930000 tstamp : 0.000000 delay : 121 avail : 1927 avail_max : 2043 [pid/9901 +954 auddev_alsa.c] Wrote 1280 bytes, frames: 640 state : XRUN trigger_time: 1194622844.333930000 tstamp : 0.000000 delay : 758 avail : 1290 avail_max : 2046 2) RAT is writing audio to the playback device for while then the device just seems to go in XRUN state when there's still positive delay in the buffer. Here's last write that went ok - showing the snd_pcm_status_dump before doing the write and afterwards: [+945 auddev_alsa.c] Time since last write: 42635us, Trying to write 640 frames state : RUNNING trigger_time: 1194622843.886113000 tstamp : 0.000000 delay : 275 avail : 1773 avail_max : 2041 [+954 auddev_alsa.c] Wrote 1280 bytes, frames: 640 state : RUNNING trigger_time: 1194622843.886113000 tstamp : 0.000000 delay : 907 avail : 1141 avail_max : 2043 The next time the audio write is called the state is checked and found to be in XRUN with 225 delay in the buffer:??? (And yet I've seen other writes find the device still running with alot less delay in the buffer??): [ +936 auddev_alsa.c] In SND_PCM_STATE_XRUN - preparing audio [pid/9901 +945 auddev_alsa.c] Frames avail to be written=2048, time diff=43182, Trying to write 640 frames, Curr delay 0 state : XRUN trigger_time: 1194622843.886113000 tstamp : 0.000000 delay : 225 avail : 1823 avail_max : 2049 [pid/9901 +954 auddev_alsa.c] Wrote 1280 bytes, frames: 640 state : RUNNING trigger_time: 1194622844.142516000 tstamp : 0.000000 delay : 637 avail : 1411 avail_max : 1708 Currently I'm stumped at to solve these problems.... And basically ALSA sound quality as worse than the OSS driver for RAT as a result. Any suggestions are to how improve or sort these problems is welcome. Thanks, Piers O'Hanlon. University College London ===snd_pcm_dump_setup:=============== stream : PLAYBACK access : RW_INTERLEAVED format : S16_LE subformat : STD channels : 1 rate : 16000 exact rate : 16000 (16000/1) msbits : 16 buffer_size : 2048 period_size : 341 period_time : 21333 tick_time : 0 tstamp_mode : NONE period_step : 1 sleep_min : 0 avail_min : 341 xfer_align : 1 start_threshold : 341 stop_threshold : 2048 silence_threshold: 0 silence_size : 0 boundary : 536870912

2 3

Re: [alsa-devel] [linux-usb-devel] [PATCH] ehci-hcd: complete iso urbs ASAP for number_of_packets != (n * 8) also
by Karsten Wiese 09 Nov '07

09 Nov '07

Am Donnerstag, 8. November 2007 schrieb David Brownell: > > Looks interesting, but could you make this apply to *BOTH* types of > ISO transfer descriptor? You did it for ITDs, but not Split-ITDs. Done. Split-ITD changes untested. > > Plus there are two non-technical problems with this patch. First, > it doesn't apply against 2.6.24-rc2-git; second, it triggers lots > of messages from scripts/checkpatch.pl ... both issues need to get > fixed before I can properly review this. Fixed. Karsten -------------------------------------> From: Karsten Wiese <fzu(a)wemgehoertderstaat.de> ehci-hcd: complete iso urbs ASAP for number_of_packets != (n * 8) also Upstream ehci-hcd only completes iso urbs, if the last frame (= 8 uframes) they overlapped with has elapsed. That can be as late as when the following urb emits its interrupt on completion. soundcard drivers tend to work around by only transfering iso urbs with number_of_packets = (n * 8). Patch lets iso urbs complete asap by scanning ITDs always up to the elapsed uframe. An ITD's last scanned uframe is stored in the new struct ehci_itd member uframe_scanned. ITDs stay kept in the hcd's schedule until the complete frame they cover has elapsed, like without patch. To make this possible, ehci->periodic_sched is changed based on the number of active itds instead of the number of active iso urbs. SITD handling is unchanged, only sitd_complete() is renamed to sitd_scan() and the active check is moved to sitd_scan() to make the "case Q_TYPE_ITD:" and "case Q_TYPE_SITD:" sections under scan_periodic() look alike. Signed-off-by: Karsten Wiese <fzu(a)wemgehoertderstaat.de> --- drivers/usb/host/ehci-sched.c | 156 ++++++++++++++++++++--------------------- drivers/usb/host/ehci.h | 3 +- 2 files changed, 76 insertions(+), 83 deletions(-) diff --git a/drivers/usb/host/ehci-sched.c b/drivers/usb/host/ehci-sched.c index 80d99bc..e647842 100644 --- a/drivers/usb/host/ehci-sched.c +++ b/drivers/usb/host/ehci-sched.c @@ -1486,6 +1486,8 @@ itd_link (struct ehci_hcd *ehci, unsigned frame, struct ehci_itd *itd) itd->frame = frame; wmb (); ehci->periodic[frame] = cpu_to_hc32(ehci, itd->itd_dma | Q_TYPE_ITD); + if (unlikely(!ehci->periodic_sched++)) + enable_periodic(ehci); } /* fit urb's itds into the selected schedule slot; activate as needed */ @@ -1558,30 +1560,28 @@ itd_link_urb ( urb->hcpriv = NULL; timer_action (ehci, TIMER_IO_WATCHDOG); - if (unlikely (!ehci->periodic_sched++)) - return enable_periodic (ehci); return 0; } #define ISO_ERRS (EHCI_ISOC_BUF_ERR | EHCI_ISOC_BABBLE | EHCI_ISOC_XACTERR) -static unsigned -itd_complete ( - struct ehci_hcd *ehci, - struct ehci_itd *itd -) { +static bool +itd_scan(struct ehci_hcd *ehci, struct ehci_itd *itd, unsigned uframe_after) +{ struct urb *urb = itd->urb; struct usb_iso_packet_descriptor *desc; u32 t; unsigned uframe; int urb_index = -1; struct ehci_iso_stream *stream = itd->stream; - struct usb_device *dev; /* for each uframe with a packet */ - for (uframe = 0; uframe < 8; uframe++) { + for (uframe = itd->uframe_scanned; uframe < uframe_after; uframe++) { + struct usb_device *dev; + if (likely (itd->index[uframe] == -1)) continue; + urb_index = itd->index[uframe]; desc = &urb->iso_frame_desc [urb_index]; @@ -1608,45 +1608,54 @@ itd_complete ( desc->status = 0; desc->actual_length = EHCI_ITD_LENGTH (t); } - } - usb_put_urb (urb); - itd->urb = NULL; - itd->stream = NULL; - list_move (&itd->itd_list, &stream->free_list); - iso_stream_put (ehci, stream); - - /* handle completion now? */ - if (likely ((urb_index + 1) != urb->number_of_packets)) - return 0; + /* handle completion now? */ + if (urb_index + 1 != urb->number_of_packets) + continue; - /* ASSERT: it's really the last itd for this urb - list_for_each_entry (itd, &stream->td_list, itd_list) - BUG_ON (itd->urb == urb); - */ + usb_put_urb(urb); + /* ASSERT: it's really the last itd for this urb + struct ehci_itd *stritd; + list_for_each_entry(stritd, &stream->td_list, itd_list) + BUG_ON(stritd != itd && stritd->urb == urb); + */ + itd->urb = NULL; - /* give urb back to the driver ... can be out-of-order */ - dev = urb->dev; - ehci_urb_done(ehci, urb, 0); - urb = NULL; + /* give urb back to the driver ... can be out-of-order */ + dev = urb->dev; + ehci_urb_done(ehci, urb, 0); + urb = NULL; - /* defer stopping schedule; completion can submit */ - ehci->periodic_sched--; - if (unlikely (!ehci->periodic_sched)) - (void) disable_periodic (ehci); - ehci_to_hcd(ehci)->self.bandwidth_isoc_reqs--; + ehci_to_hcd(ehci)->self.bandwidth_isoc_reqs--; - if (unlikely (list_empty (&stream->td_list))) { - ehci_to_hcd(ehci)->self.bandwidth_allocated + if (unlikely(list_empty(&stream->td_list))) { + ehci_to_hcd(ehci)->self.bandwidth_allocated -= stream->bandwidth; - ehci_vdbg (ehci, - "deschedule devp %s ep%d%s-iso\n", - dev->devpath, stream->bEndpointAddress & 0x0f, - (stream->bEndpointAddress & USB_DIR_IN) ? "in" : "out"); + ehci_vdbg(ehci, + "deschedule devp %s ep%d%s-iso\n", + dev->devpath, + stream->bEndpointAddress & 0x0f, + (stream->bEndpointAddress & USB_DIR_IN) ? + "in" : "out"); + } + iso_stream_put(ehci, stream); } + + itd->uframe_scanned = uframe_after; + if (uframe_after < 8) + return false; + + itd->stream = NULL; + itd->uframe_scanned = 0; + list_move(&itd->itd_list, &stream->free_list); iso_stream_put (ehci, stream); - return 1; + /* defer stopping schedule; completion can submit */ + ehci->periodic_sched--; + if (unlikely(!ehci->periodic_sched)) + disable_periodic(ehci); + + return true; } /*-------------------------------------------------------------------------*/ @@ -1950,11 +1959,9 @@ sitd_link_urb ( #define SITD_ERRS (SITD_STS_ERR | SITD_STS_DBE | SITD_STS_BABBLE \ | SITD_STS_XACT | SITD_STS_MMF) -static unsigned -sitd_complete ( - struct ehci_hcd *ehci, - struct ehci_sitd *sitd -) { +static bool +sitd_scan(struct ehci_hcd *ehci, struct ehci_sitd *sitd, int live) +{ struct urb *urb = sitd->urb; struct usb_iso_packet_descriptor *desc; u32 t; @@ -1962,9 +1969,12 @@ sitd_complete ( struct ehci_iso_stream *stream = sitd->stream; struct usb_device *dev; + t = hc32_to_cpup(ehci, &sitd->hw_results); + if (t & SITD_STS_ACTIVE && live) + return false; + urb_index = sitd->index; desc = &urb->iso_frame_desc [urb_index]; - t = hc32_to_cpup(ehci, &sitd->hw_results); /* report transfer status */ if (t & SITD_ERRS) { @@ -1991,7 +2001,7 @@ sitd_complete ( /* handle completion now? */ if ((urb_index + 1) != urb->number_of_packets) - return 0; + return true; /* ASSERT: it's really the last sitd for this urb list_for_each_entry (sitd, &stream->td_list, sitd_list) @@ -2019,7 +2029,7 @@ sitd_complete ( } iso_stream_put (ehci, stream); - return 1; + return true; } @@ -2091,13 +2101,11 @@ sitd_submit (struct ehci_hcd *ehci, struct urb *urb, gfp_t mem_flags) return -ENOSYS; } -static inline unsigned -sitd_complete ( - struct ehci_hcd *ehci, - struct ehci_sitd *sitd -) { - ehci_err (ehci, "sitd_complete %p?\n", sitd); - return 0; +static inline bool +sitd_scan(struct ehci_hcd *ehci, struct ehci_sitd *sitd) +{ + ehci_err(ehci, "%s %p?\n", __FUNCTION__, sitd); + return false; } #endif /* USB_EHCI_SPLIT_ISO */ @@ -2175,47 +2183,33 @@ restart: q = q.fstn->fstn_next; break; case Q_TYPE_ITD: - /* skip itds for later in the frame */ rmb (); - for (uf = live ? uframes : 8; uf < 8; uf++) { - if (0 == (q.itd->hw_transaction [uf] - & ITD_ACTIVE(ehci))) - continue; + modified = itd_scan(ehci, q.itd, + uf = live ? uframes : 8); + if (!modified) { q_p = &q.itd->itd_next; hw_p = &q.itd->hw_next; - type = Q_NEXT_TYPE(ehci, - q.itd->hw_next); - q = *q_p; - break; + } else { + *q_p = q.itd->itd_next; + *hw_p = q.itd->hw_next; } - if (uf != 8) - break; - - /* this one's ready ... HC won't cache the - * pointer for much longer, if at all. - */ - *q_p = q.itd->itd_next; - *hw_p = q.itd->hw_next; type = Q_NEXT_TYPE(ehci, q.itd->hw_next); + wmb(); - modified = itd_complete (ehci, q.itd); q = *q_p; break; case Q_TYPE_SITD: - if ((q.sitd->hw_results & SITD_ACTIVE(ehci)) - && live) { + modified = sitd_scan(ehci, q.sitd, live); + if (!modified) { q_p = &q.sitd->sitd_next; hw_p = &q.sitd->hw_next; - type = Q_NEXT_TYPE(ehci, - q.sitd->hw_next); - q = *q_p; - break; + } else { + *q_p = q.sitd->sitd_next; + *hw_p = q.sitd->hw_next; } - *q_p = q.sitd->sitd_next; - *hw_p = q.sitd->hw_next; type = Q_NEXT_TYPE(ehci, q.sitd->hw_next); + wmb(); - modified = sitd_complete (ehci, q.sitd); q = *q_p; break; default: diff --git a/drivers/usb/host/ehci.h b/drivers/usb/host/ehci.h index 951d69f..156c02f 100644 --- a/drivers/usb/host/ehci.h +++ b/drivers/usb/host/ehci.h @@ -581,6 +581,7 @@ struct ehci_itd { struct urb *urb; struct ehci_iso_stream *stream; /* endpoint's queue */ struct list_head itd_list; /* list of stream's itds */ + unsigned uframe_scanned; /* any/all hw_transactions here may be used by that urb */ unsigned frame; /* where scheduled */ @@ -615,8 +616,6 @@ struct ehci_sitd { #define SITD_STS_MMF (1 << 2) /* incomplete split transaction */ #define SITD_STS_STS (1 << 1) /* split transaction state */ -#define SITD_ACTIVE(ehci) cpu_to_hc32(ehci, SITD_STS_ACTIVE) - __hc32 hw_buf [2]; /* EHCI table 3-12 */ __hc32 hw_backpointer; /* EHCI table 3-13 */ __hc32 hw_buf_hi [2]; /* Appendix B */ -- 1.5.3.3

1 0

[alsa-devel] [PATCH] sound - remove dead config symbol from sound code
by Jiri Olsa 09 Nov '07

09 Nov '07

remove dead config symbols from sound code Signed-off-by: Jiri Olsa <olsajiri(a)gmail.com> --- include/sound/cs46xx.h | 3 - include/sound/driver.h | 4 -- sound/pci/cs46xx/cs46xx_lib.c | 99 ----------------------------------------- 3 files changed, 0 insertions(+), 106 deletions(-) diff --git a/include/sound/cs46xx.h b/include/sound/cs46xx.h index 6b40ee6..e3005a6 100644 --- a/include/sound/cs46xx.h +++ b/include/sound/cs46xx.h @@ -1708,9 +1708,6 @@ struct snd_cs46xx { struct gameport *gameport; -#ifdef CONFIG_SND_CS46XX_DEBUG_GPIO - int current_gpio; -#endif #ifdef CONFIG_SND_CS46XX_NEW_DSP struct mutex spos_mutex; diff --git a/include/sound/driver.h b/include/sound/driver.h index 5ccb6c5..1889929 100644 --- a/include/sound/driver.h +++ b/include/sound/driver.h @@ -38,10 +38,6 @@ #define CONFIG_SND_MAJOR 116 #endif -#ifndef CONFIG_SND_DEBUG -#undef CONFIG_SND_DEBUG_MEMORY -#endif - #ifdef ALSA_BUILD #include "adriver.h" #endif diff --git a/sound/pci/cs46xx/cs46xx_lib.c b/sound/pci/cs46xx/cs46xx_lib.c index 2c7bfc9..8c44fef 100644 --- a/sound/pci/cs46xx/cs46xx_lib.c +++ b/sound/pci/cs46xx/cs46xx_lib.c @@ -2084,71 +2084,6 @@ static int snd_cs46xx_spdif_stream_put(struct snd_kcontrol *kcontrol, #endif /* CONFIG_SND_CS46XX_NEW_DSP */ -#ifdef CONFIG_SND_CS46XX_DEBUG_GPIO -static int snd_cs46xx_egpio_select_info(struct snd_kcontrol *kcontrol, - struct snd_ctl_elem_info *uinfo) -{ - uinfo->type = SNDRV_CTL_ELEM_TYPE_INTEGER; - uinfo->count = 1; - uinfo->value.integer.min = 0; - uinfo->value.integer.max = 8; - return 0; -} - -static int snd_cs46xx_egpio_select_get(struct snd_kcontrol *kcontrol, - struct snd_ctl_elem_value *ucontrol) -{ - struct snd_cs46xx *chip = snd_kcontrol_chip(kcontrol); - ucontrol->value.integer.value[0] = chip->current_gpio; - - return 0; -} - -static int snd_cs46xx_egpio_select_put(struct snd_kcontrol *kcontrol, - struct snd_ctl_elem_value *ucontrol) -{ - struct snd_cs46xx *chip = snd_kcontrol_chip(kcontrol); - int change = (chip->current_gpio != ucontrol->value.integer.value[0]); - chip->current_gpio = ucontrol->value.integer.value[0]; - - return change; -} - - -static int snd_cs46xx_egpio_get(struct snd_kcontrol *kcontrol, - struct snd_ctl_elem_value *ucontrol) -{ - struct snd_cs46xx *chip = snd_kcontrol_chip(kcontrol); - int reg = kcontrol->private_value; - - snd_printdd ("put: reg = %04x, gpio %02x\n",reg,chip->current_gpio); - ucontrol->value.integer.value[0] = - (snd_cs46xx_peekBA0(chip, reg) & (1 << chip->current_gpio)) ? 1 : 0; - - return 0; -} - -static int snd_cs46xx_egpio_put(struct snd_kcontrol *kcontrol, - struct snd_ctl_elem_value *ucontrol) -{ - struct snd_cs46xx *chip = snd_kcontrol_chip(kcontrol); - int reg = kcontrol->private_value; - int val = snd_cs46xx_peekBA0(chip, reg); - int oldval = val; - snd_printdd ("put: reg = %04x, gpio %02x\n",reg,chip->current_gpio); - - if (ucontrol->value.integer.value[0]) - val |= (1 << chip->current_gpio); - else - val &= ~(1 << chip->current_gpio); - - snd_cs46xx_pokeBA0(chip, reg,val); - snd_printdd ("put: val %08x oldval %08x\n",val,oldval); - - return (oldval != val); -} -#endif /* CONFIG_SND_CS46XX_DEBUG_GPIO */ - static struct snd_kcontrol_new snd_cs46xx_controls[] __devinitdata = { { .iface = SNDRV_CTL_ELEM_IFACE_MIXER, @@ -2241,40 +2176,6 @@ static struct snd_kcontrol_new snd_cs46xx_controls[] __devinitdata = { }, #endif -#ifdef CONFIG_SND_CS46XX_DEBUG_GPIO -{ - .iface = SNDRV_CTL_ELEM_IFACE_MIXER, - .name = "EGPIO select", - .info = snd_cs46xx_egpio_select_info, - .get = snd_cs46xx_egpio_select_get, - .put = snd_cs46xx_egpio_select_put, - .private_value = 0, -}, -{ - .iface = SNDRV_CTL_ELEM_IFACE_MIXER, - .name = "EGPIO Input/Output", - .info = snd_mixer_boolean_info, - .get = snd_cs46xx_egpio_get, - .put = snd_cs46xx_egpio_put, - .private_value = BA0_EGPIODR, -}, -{ - .iface = SNDRV_CTL_ELEM_IFACE_MIXER, - .name = "EGPIO CMOS/Open drain", - .info = snd_mixer_boolean_info, - .get = snd_cs46xx_egpio_get, - .put = snd_cs46xx_egpio_put, - .private_value = BA0_EGPIOPTR, -}, -{ - .iface = SNDRV_CTL_ELEM_IFACE_MIXER, - .name = "EGPIO On/Off", - .info = snd_mixer_boolean_info, - .get = snd_cs46xx_egpio_get, - .put = snd_cs46xx_egpio_put, - .private_value = BA0_EGPIOSR, -}, -#endif }; #ifdef CONFIG_SND_CS46XX_NEW_DSP

1 0

[alsa-devel] When command SNDRV_PCM_TRIGGER_STOP is asked in snd_pcm_ops trigger(), which buffer in the buffer list should be the last one to output by the DMA?
by Sonic Zhang 09 Nov '07

09 Nov '07

In our blackfin ad1981 linux driver, there is an output ring with "periods" buffers. Each buffer is of the size of "period_size". After ALSA starts playing, several buffers ahead of current one running by the DMA controller are usually filled in advance by the ALSA layer. When command SNDRV_PCM_TRIGGER_STOP is asked in snd_pcm_ops trigger(), which buffer should be the last one to output by the DMA? 1. Current buffer, which is running by the DMA controller 2. The last buffer, which the DMA controller just finished outputting and snd_pcm_ops pointer() is notifiyed to ALSA layer. Does this mean current buffer in DMA should be abandoned in the middle? 3. The last buffer, which the ALSA layer has just filled in snd_pcm_ops copy()? Thanks

2 1

[alsa-devel] User-space OSS emulation?
by Timur Tabi 09 Nov '07

09 Nov '07

Can someone tell me more about "user-space OSS emulation"? I've read a few references to it in the mailing list archives, but a google search doesn't turn up anything useful. -- Timur Tabi Linux kernel developer at Freescale

3 4

[alsa-devel] How can OSS applications support multiple streams?
by Timur Tabi 08 Nov '07

08 Nov '07

I've written an ALSA driver that supports only one playback stream, because that's what our hardware supports. I have a customer that has an OSS application that tries to open 32 simultaneous playback streams, which obviously doesn't work. What do I (or the customer) need to do to get 32 streams to work? Is this what a sound server is for? If so, which one should I use? -- Timur Tabi Linux kernel developer at Freescale

1 0

[alsa-devel] Any OSS changes from kernel 2.6.21 to 2.6.23? Something broke.
by Timur Tabi 08 Nov '07

08 Nov '07

I have a customer who says that mplayer (using OSS) does not work when he tries a kernel based on 2.6.23-rc4, but it works when he uses a kernel based on 2.6.21. Everything else is the same (including the version of alsa-lib), so I presume something broke in the OSS emulation in the kernel. Unfortunately, our git repositories for these trees are not compatible, so I can't use git-bisect to narrow the problem down. Is it possible to back-level the ALSA support in the kernel itself? Can I just download ftp://ftp.alsa-project.org/pub/driver/alsa-driver-1.0.XX.tar.bz2 and follow the instructions in the INSTALL file? If you update the kernel, are you supposed to use the latest alsa-lib as well? For a given kernel, how do I know what the right version of alsa-lib is? -- Timur Tabi Linux kernel developer at Freescale

4 9

[alsa-devel] Fwd: [PATCH] ehci-hcd: complete iso urbs ASAP for number_of_packets != (n * 8) also
by Karsten Wiese 08 Nov '07

08 Nov '07

Hi I just sent attached patch to usb-devel. It makes special workaround for iso usb communication using ehci-hcd highspeed needless. Comments, Bug/success reports welcome! kind regards Karsten Upstream (as of 2.6.23) ehci-hcd only completes iso urbs, if the last frame (= 8 uframes) they overlapped with has elapsed. That can be as late as when the following urb emits its interrupt on completion. soundcard drivers tend to work around by only transfering iso urbs with number_of_packets = (n * 8). Patch lets iso urbs complete asap by scanning itd always up to the elapsed uframe. An itd's last scanned uframe is stored in the new struct ehci_itd member uframe_scanned. Itds stay kept in the hcd's schedule until the complete frame they cover has elapsed, like without patch. To make this possible, ehci->periodic_sched is changed based on the number of active itds instead of the number of active iso urbs. Signed-off-by: Karsten Wiese <fzu(a)wemgehoertderstaat.de> --- drivers/usb/host/ehci-sched.c | 108 ++++++++++++++++++++--------------------- drivers/usb/host/ehci.h | 1 + 2 files changed, 54 insertions(+), 55 deletions(-) diff --git a/drivers/usb/host/ehci-sched.c b/drivers/usb/host/ehci-sched.c index e682f23..c456b16 100644 --- a/drivers/usb/host/ehci-sched.c +++ b/drivers/usb/host/ehci-sched.c @@ -1481,6 +1481,8 @@ itd_link (struct ehci_hcd *ehci, unsigned frame, struct ehci_itd *itd) itd->frame = frame; wmb (); ehci->periodic[frame] = cpu_to_hc32(ehci, itd->itd_dma | Q_TYPE_ITD); + if (unlikely (!ehci->periodic_sched++)) + enable_periodic (ehci); } /* fit urb's itds into the selected schedule slot; activate as needed */ @@ -1553,30 +1555,30 @@ itd_link_urb ( urb->hcpriv = NULL; timer_action (ehci, TIMER_IO_WATCHDOG); - if (unlikely (!ehci->periodic_sched++)) - return enable_periodic (ehci); return 0; } #define ISO_ERRS (EHCI_ISOC_BUF_ERR | EHCI_ISOC_BABBLE | EHCI_ISOC_XACTERR) -static unsigned -itd_complete ( - struct ehci_hcd *ehci, - struct ehci_itd *itd -) { +static bool +itd_scan ( + struct ehci_hcd *ehci, + struct ehci_itd *itd, + unsigned uframe_after +) +{ struct urb *urb = itd->urb; struct usb_iso_packet_descriptor *desc; u32 t; unsigned uframe; int urb_index = -1; struct ehci_iso_stream *stream = itd->stream; - struct usb_device *dev; /* for each uframe with a packet */ - for (uframe = 0; uframe < 8; uframe++) { + for (uframe = itd->uframe_scanned; uframe < uframe_after; uframe++) { if (likely (itd->index[uframe] == -1)) continue; + urb_index = itd->index[uframe]; desc = &urb->iso_frame_desc [urb_index]; @@ -1603,45 +1605,53 @@ itd_complete ( desc->status = 0; desc->actual_length = EHCI_ITD_LENGTH (t); } + + /* handle completion now? */ + if (urb_index + 1 == urb->number_of_packets) { + struct usb_device *dev; + + usb_put_urb (urb); + /* ASSERT: it's really the last itd for this urb + struct ehci_itd *stritd; + list_for_each_entry (stritd, &stream->td_list, itd_list) + BUG_ON (stritd != itd && stritd->urb == urb); + */ + itd->urb = NULL; + + /* give urb back to the driver ... can be out-of-order */ + dev = urb->dev; + ehci_urb_done (ehci, urb); + urb = NULL; + + ehci_to_hcd(ehci)->self.bandwidth_isoc_reqs--; + + if (unlikely (list_empty (&stream->td_list))) { + ehci_to_hcd(ehci)->self.bandwidth_allocated + -= stream->bandwidth; + ehci_vdbg (ehci, + "deschedule devp %s ep%d%s-iso\n", + dev->devpath, stream->bEndpointAddress & 0x0f, + (stream->bEndpointAddress & USB_DIR_IN) ? "in" : "out"); + } + iso_stream_put (ehci, stream); + } } - usb_put_urb (urb); - itd->urb = NULL; + itd->uframe_scanned = uframe_after; + if (uframe_after < 8) + return false; + itd->stream = NULL; + itd->uframe_scanned = 0; list_move (&itd->itd_list, &stream->free_list); iso_stream_put (ehci, stream); - /* handle completion now? */ - if (likely ((urb_index + 1) != urb->number_of_packets)) - return 0; - - /* ASSERT: it's really the last itd for this urb - list_for_each_entry (itd, &stream->td_list, itd_list) - BUG_ON (itd->urb == urb); - */ - - /* give urb back to the driver ... can be out-of-order */ - dev = urb->dev; - ehci_urb_done (ehci, urb); - urb = NULL; - /* defer stopping schedule; completion can submit */ ehci->periodic_sched--; if (unlikely (!ehci->periodic_sched)) (void) disable_periodic (ehci); - ehci_to_hcd(ehci)->self.bandwidth_isoc_reqs--; - if (unlikely (list_empty (&stream->td_list))) { - ehci_to_hcd(ehci)->self.bandwidth_allocated - -= stream->bandwidth; - ehci_vdbg (ehci, - "deschedule devp %s ep%d%s-iso\n", - dev->devpath, stream->bEndpointAddress & 0x0f, - (stream->bEndpointAddress & USB_DIR_IN) ? "in" : "out"); - } - iso_stream_put (ehci, stream); - - return 1; + return true; } /*-------------------------------------------------------------------------*/ @@ -2156,30 +2166,18 @@ restart: q = q.fstn->fstn_next; break; case Q_TYPE_ITD: - /* skip itds for later in the frame */ rmb (); - for (uf = live ? uframes : 8; uf < 8; uf++) { - if (0 == (q.itd->hw_transaction [uf] - & ITD_ACTIVE(ehci))) - continue; + modified = itd_scan(ehci, q.itd, uf = live ? uframes : 8); + if (!modified) { q_p = &q.itd->itd_next; hw_p = &q.itd->hw_next; - type = Q_NEXT_TYPE(ehci, - q.itd->hw_next); - q = *q_p; - break; + } else { + *q_p = q.itd->itd_next; + *hw_p = q.itd->hw_next; } - if (uf != 8) - break; - - /* this one's ready ... HC won't cache the - * pointer for much longer, if at all. - */ - *q_p = q.itd->itd_next; - *hw_p = q.itd->hw_next; type = Q_NEXT_TYPE(ehci, q.itd->hw_next); + wmb(); - modified = itd_complete (ehci, q.itd); q = *q_p; break; case Q_TYPE_SITD: diff --git a/drivers/usb/host/ehci.h b/drivers/usb/host/ehci.h index 951d69f..df35964 100644 --- a/drivers/usb/host/ehci.h +++ b/drivers/usb/host/ehci.h @@ -581,6 +581,7 @@ struct ehci_itd { struct urb *urb; struct ehci_iso_stream *stream; /* endpoint's queue */ struct list_head itd_list; /* list of stream's itds */ + unsigned uframe_scanned; /* any/all hw_transactions here may be used by that urb */ unsigned frame; /* where scheduled */ -- 1.5.3.3

2 6

[alsa-devel] Audio quality issues with dmix plugin
by Pharaoh . 08 Nov '07

08 Nov '07

Hi, When I play a 44Khz file without using any ALSA plugin like #aplay -Dhw:0,0 44k.wav the sound quality is awesome but when I play the same file using dmix, clicks are introduced. #aplay -Dplug:dmix 44k.wav changing period size and buffer size doesn't help. I am testing my ALSA driver on an OMAP850 based board. Regards, Pharaoh.

2 4