intel8x0: [v2] Improve performance in virtual environment
A message that is not in the archive (sent off-list or not yet received): [email protected]
At Wed, 26 Oct 2011 16:00:15 +0400,
Quoted 24 lines · Konstantin Ozerkov
v2: add detection for virtual environments (KVM and Parallels)
This patch is intended to improve performance in virtualized environments like Parallels Desktop or KVM/VirtualBox/QEMU (virtual ICH/AC97 audio).
I/O access is very time-expensive operation in virtual world: VCPU can be rescheduled and in the worst case we get more than 10ms delay on each I/O access.
In the virtual environment loop exit rule
(old_civ==current_civ&&old_picb==current_picb) is never satisfied, becauseold_picbis never the same ascurrent_picbdue to delay inspired by readingcurrent_civ. As a result loop ended by timeout and we get 10x more I/O operations.Experimental data from Prallels Desktop 7, RHEL6 guest (I/O ops per second):
Original code:
In Port Counter Callback f014 41550 fffff00000179d00 ac97_bm_read_civ+0x000 f018 41387 fffff0000017a580 ac97_bm_read_picb+0x000
Quoted 7 lines
With patch:
In Port Counter Callback f014 4090 fffff00000179d00 ac97_bm_read_civ+0x000 f018 1964 fffff0000017a580 ac97_bm_read_picb+0x000Signed-off-by Konstantin Ozerkov <[email protected]>Signed-off-by Denis V. Lunev <[email protected]>
Are the changes specific to x86, right?boot_cpu_has() seems x86 only, and kvm_para_*() isn't provided for all architectures.
The best would be to provide a standard API for such detection. But, as a workaround, we may need a simple ifdef.
thanks,
Takashi
Quoted 77 lines
pci/intel8x0.c +19
-2@@ -27,6 +27,7 @@*/ #include <asm/io.h> +#include <asm/kvm_para.h> #include <linux/delay.h> #include <linux/interrupt.h> #include <linux/init.h> @@ -77,6 +78,7 @@ static int buggy_semaphore;static int buggy_irq =-1 ;/* auto-check */ static int xbox;static int spdif_aclink =-1 ; +static int inside_vm =-1 ;module_param (index, int,0444 );MODULE_PARM_DESC (index,"Index value for Intel i8x0 soundcard." );@@ -94,6 +96,8 @@ module_param(xbox, bool, 0444);MODULE_PARM_DESC (xbox,"Set to 1 for Xbox, if you have problems with the AC'97 codec detection." );module_param (spdif_aclink, int,0444 );MODULE_PARM_DESC (spdif_aclink,"S/PDIF over AC-link." ); +module_param (inside_vm, bool,0444 ); +MODULE_PARM_DESC (inside_vm,"KVM/Parallels optimization." );/* just for backward compatibility */ static int enable;@@ -400,6 +404,7 @@ struct intel8x0 { unsigned buggy_irq: 1; /* workaround for buggy mobos */ unsigned xbox: 1; /* workaround for Xbox AC'97 detection */ unsigned buggy_semaphore: 1; /* workaround for buggy codec semaphore */ + unsigned inside_vm: 1; /* enable VM optimization */int spdif_idx;/* SPDIF BAR index; *_SPBAR or -1 if use PCMOUT */ unsigned int sdm_saved;/* SDM reg value */ @@ -1065,8 +1070,11 @@ static snd_pcm_uframes_t snd_intel8x0_pcm_pointer(struct snd_pcm_substream *subsudelay (10 );continue ; } -if (civ ==igetbyte (chip, ichdev->reg_offset +ICH_REG_OFF_CIV ) && - ptr1 ==igetword (chip, ichdev->reg_offset + ichdev->roff_picb)) +if (civ !=igetbyte (chip, ichdev->reg_offset +ICH_REG_OFF_CIV )) +continue ; +if (chip->inside_vm) +break ; +if (ptr1 ==igetword (chip, ichdev->reg_offset + ichdev->roff_picb))break ; }while (timeout--); ptr = ichdev->last_pos;@@ -2984,6 +2992,10 @@ static int __devinit snd_intel8x0_create(struct snd_card *card,if (xbox) chip->xbox =1 ; + chip->inside_vm = inside_vm; +if (inside_vm) +printk (KERN_INFO "intel8x0: enable KVM optimization\n" ); +if (pci->vendor ==PCI_VENDOR_ID_INTEL && pci->device ==PCI_DEVICE_ID_INTEL_440MX ) chip->fix_nocache =1 ;/* enable workaround */ @@ -3226,6 +3238,11 @@ static int __devinit snd_intel8x0_probe(struct pci_dev *pci, buggy_irq =0 ; } +if (inside_vm <0 ) { +/* detect KVM and Parallels virtual environments */ + inside_vm =kvm_para_available () ||boot_cpu_has (X86_FEATURE_HYPERVISOR ); + } +if ((err =snd_intel8x0_create (card, pci, pci_id->driver_data, &chip)) <0 ) {snd_card_free (card);Signature · 1 line
1.7.7
A message that is not in the archive (sent off-list or not yet received): [email protected]
At Wed, 26 Oct 2011 17:29:37 +0400,
Quoted 9 lines · Konstantin Ozerkov
Quoted 6 lines · Takashi Iwai
Are the changes specific to x86, right?
boot_cpu_has()seems x86 only, and kvm_para_*() isn't provided for all architectures.The best would be to provide a standard API for such detection. But, as a workaround, we may need a simple ifdef.
Is this really needed ? This is driver for controllers which is used only (AFIK) with x86 and
x86_64CPUs.
I remember vaguely the implementation on PPC and others. Who knows. (And you can run a kernel in VM, right? :)
Takashi