[alsa-devel] [PATCH 0/4] More hdspm fixes
Hi!
Here are four more patches to hdspm. Most importantly, they add the missing PCI firmware revision for AES32. Without it, loading hdspm would crash on machines with these cards. To prevent such a crash upon unknown firmware revisions, I've created a second patch adding a default case to the firmware switch-case statement.
Also worth mentioning: midi port initialization is now no longer hardcoded, thus fixing a bootup problem on MADIface.
All patches have been runtime tested on MADIface, AES32, AES (PCIe) and RayDAT.
Cheers
Adrian Knoth (4): ALSA: hdspm - AES32: Use define instead of hardcoded channel count ALSA: hdspm - Add updated firmware revision for AES32 ALSA: hdspm - Abort on unknown firmware revision ALSA: hdspm - Fix midi port initialization
sound/pci/rme9652/hdspm.c | 24 ++++++++++++++++++------ 1 files changed, 18 insertions(+), 6 deletions(-)
Signed-off-by: Adrian Knoth adi@drcomp.erfurt.thur.de
diff --git a/sound/pci/rme9652/hdspm.c b/sound/pci/rme9652/hdspm.c index 5dbf620..7ce4929 100644 --- a/sound/pci/rme9652/hdspm.c +++ b/sound/pci/rme9652/hdspm.c @@ -506,6 +506,8 @@ MODULE_SUPPORTED_DEVICE("{{RME HDSPM-MADI}}"); #define AIO_OUT_DS_CHANNELS 12 #define AIO_OUT_QS_CHANNELS 10
+#define AES32_CHANNELS 16 + /* the size of a substream (1 mono data stream) */ #define HDSPM_CHANNEL_BUFFER_SAMPLES (16*1024) #define HDSPM_CHANNEL_BUFFER_BYTES (4*HDSPM_CHANNEL_BUFFER_SAMPLES) @@ -6449,9 +6451,9 @@ static int __devinit snd_hdspm_create(struct snd_card *card,
switch (hdspm->io_type) { case AES32: - hdspm->ss_in_channels = hdspm->ss_out_channels = 16; - hdspm->ds_in_channels = hdspm->ds_out_channels = 16; - hdspm->qs_in_channels = hdspm->qs_out_channels = 16; + hdspm->ss_in_channels = hdspm->ss_out_channels = AES32_CHANNELS; + hdspm->ds_in_channels = hdspm->ds_out_channels = AES32_CHANNELS; + hdspm->qs_in_channels = hdspm->qs_out_channels = AES32_CHANNELS;
hdspm->channel_map_in_ss = hdspm->channel_map_out_ss = channel_map_aes32; @@ -6466,7 +6468,8 @@ static int __devinit snd_hdspm_create(struct snd_card *card, hdspm->port_names_in_qs = hdspm->port_names_out_qs = texts_ports_aes32;
- hdspm->max_channels_out = hdspm->max_channels_in = 16; + hdspm->max_channels_out = hdspm->max_channels_in = + AES32_CHANNELS; hdspm->port_names_in = hdspm->port_names_out = texts_ports_aes32; hdspm->channel_map_in = hdspm->channel_map_out =
In contrast to the PCIe version (RME AES), the PCI version (RME AES32) has a different firmware revision.
This patch adds the missing PCI revision.
Signed-off-by: Adrian Knoth adi@drcomp.erfurt.thur.de
diff --git a/sound/pci/rme9652/hdspm.c b/sound/pci/rme9652/hdspm.c index 7ce4929..20c54f7 100644 --- a/sound/pci/rme9652/hdspm.c +++ b/sound/pci/rme9652/hdspm.c @@ -526,6 +526,7 @@ MODULE_SUPPORTED_DEVICE("{{RME HDSPM-MADI}}"); #define HDSPM_AIO_REV 212 #define HDSPM_MADIFACE_REV 213 #define HDSPM_AES_REV 240 +#define HDSPM_AES32_REV 234
/* speed factor modes */ #define HDSPM_SPEED_SINGLE 0 @@ -6393,6 +6394,7 @@ static int __devinit snd_hdspm_create(struct snd_card *card, hdspm->midiPorts = 1; break; case HDSPM_AES_REV: + case HDSPM_AES32_REV: hdspm->io_type = AES32; hdspm->card_name = "RME AES32"; hdspm->midiPorts = 2;
Don't continue if we cannot detect the real card type, otherwise, all subsequent functions, especially strcpy(), would fail, leaving the whole driver in an unusable state.
Without such a protection, dmesg would look like this:
Pid: 1525, comm: modprobe Not tainted 2.6.38-rc6 #1 MSI MS-7250/MS-7250 EIP: 0060:[<c114e700>] EFLAGS: 00010296 CPU: 1 EIP is at strcpy+0x10/0x30 EAX: f4d33e58 EBX: f5990800 ECX: f4d33e58 EDX: 00000000 ESI: 00000000 EDI: f4d33e58 EBP: f5990930 ESP: f3dd3e0c DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 f53d4204 00000000 f90bfc03 00000001 00000001 f3dd3e64 c10fedff f3dd3e7c f3d854e4 f53d4000 c10307b8 00000001 c10307b8 f5990860 c10307b8 00000001 c10332e5 f59908bc c12f4339 f59908bc f5990860 c11ef207 f4d33e00 f53d4000 [<f90bfc03>] ? snd_hdspm_probe+0x7e7/0x1166 [snd_hdspm] [<c10fedff>] ? __sysfs_add_one+0x1f/0xf0 [<c10307b8>] ? get_parent_ip+0x8/0x20 [<c10307b8>] ? get_parent_ip+0x8/0x20 [<c10307b8>] ? get_parent_ip+0x8/0x20 [<c10332e5>] ? add_preempt_count+0xa5/0xd0 [<c12f4339>] ? _raw_spin_lock_irqsave+0x19/0x40 [<c11ef207>] ? pm_runtime_enable+0x17/0x80 [<c1160172>] ? local_pci_probe+0x42/0xb0 [<c1161091>] ? pci_device_probe+0x61/0x80 [<c11e9587>] ? driver_probe_device+0x77/0x180 [<c11600f0>] ? pci_match_device+0xa0/0xc0 [<c11e9709>] ? __driver_attach+0x79/0x80 [<c11e9690>] ? __driver_attach+0x0/0x80 [<c11e8e32>] ? bus_for_each_dev+0x52/0x80 [<c11e9416>] ? driver_attach+0x16/0x20 [<c11e9690>] ? __driver_attach+0x0/0x80 [<c11e8771>] ? bus_add_driver+0xa1/0x220 [<c1160f20>] ? pci_device_remove+0x0/0xf0 [<c11e99a3>] ? driver_register+0x63/0x120 [<c11612bd>] ? __pci_register_driver+0x3d/0xb0 [<c1001132>] ? do_one_initcall+0x32/0x160 [<f90d6000>] ? alsa_card_hdspm_init+0x0/0x14 [snd_hdspm] [<c10676a9>] ? sys_init_module+0x99/0x1e0 [<c10ade3d>] ? sys_close+0x6d/0xc0 [<c1002c90>] ? sysenter_do_call+0x12/0x26 ---[ end trace 239f0661c558378b ]---
Signed-off-by: Adrian Knoth adi@drcomp.erfurt.thur.de
diff --git a/sound/pci/rme9652/hdspm.c b/sound/pci/rme9652/hdspm.c index 20c54f7..78f6ab5 100644 --- a/sound/pci/rme9652/hdspm.c +++ b/sound/pci/rme9652/hdspm.c @@ -6399,6 +6399,10 @@ static int __devinit snd_hdspm_create(struct snd_card *card, hdspm->card_name = "RME AES32"; hdspm->midiPorts = 2; break; + default: + snd_printk(KERN_ERR "HDSPM: unknown firmware revision %x\n", + hdspm->firmware_rev); + return -ENODEV; }
err = pci_enable_device(pci);
Depending on the model and the presence of a TCO module, the number of midi ports varies. Some have 1 port (MADIface), some have 2 (default), with TCO, there are 3.
Don't hardcode the number of midi ports to initialize.
This patch also fixes a boot lockup on MADIface.
Signed-off-by: Adrian Knoth adi@drcomp.erfurt.thur.de
diff --git a/sound/pci/rme9652/hdspm.c b/sound/pci/rme9652/hdspm.c index 78f6ab5..3224766 100644 --- a/sound/pci/rme9652/hdspm.c +++ b/sound/pci/rme9652/hdspm.c @@ -6291,8 +6291,11 @@ static int __devinit snd_hdspm_create_pcm(struct snd_card *card,
static inline void snd_hdspm_initialize_midi_flush(struct hdspm * hdspm) { - snd_hdspm_flush_midi_input(hdspm, 0); - snd_hdspm_flush_midi_input(hdspm, 1); + int i; + + for (i=0; i < hdspm->midiPorts; i++) { + snd_hdspm_flush_midi_input(hdspm, i); + } }
static int __devinit snd_hdspm_create_alsa_devices(struct snd_card *card,
At Mon, 28 Feb 2011 15:14:46 +0100, Adrian Knoth wrote:
Hi!
Here are four more patches to hdspm. Most importantly, they add the missing PCI firmware revision for AES32. Without it, loading hdspm would crash on machines with these cards. To prevent such a crash upon unknown firmware revisions, I've created a second patch adding a default case to the firmware switch-case statement.
Also worth mentioning: midi port initialization is now no longer hardcoded, thus fixing a bootup problem on MADIface.
All patches have been runtime tested on MADIface, AES32, AES (PCIe) and RayDAT.
Thanks. Applied now (with a slight coding-style fix).
Takashi
Cheers
Adrian Knoth (4): ALSA: hdspm - AES32: Use define instead of hardcoded channel count ALSA: hdspm - Add updated firmware revision for AES32 ALSA: hdspm - Abort on unknown firmware revision ALSA: hdspm - Fix midi port initialization
sound/pci/rme9652/hdspm.c | 24 ++++++++++++++++++------ 1 files changed, 18 insertions(+), 6 deletions(-)
-- 1.7.4.1
participants (2)
-
Adrian Knoth -
Takashi Iwai