Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1452394 > unrolled thread

Re: [PATCH] usb:serial: Add Fintek F81532/534 driver

Started byOne Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk>
First post2016-07-29 14:50 +0200
Last post2016-08-04 08:40 +0200
Articles 2 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.


Contents

  Re: [PATCH] usb:serial: Add Fintek F81532/534 driver One Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk> - 2016-07-29 14:50 +0200
    Re: [PATCH] usb:serial: Add Fintek F81532/534 driver "Ji-Ze Hong (Peter Hong)" <hpeter@gmail.com> - 2016-08-04 08:40 +0200

#1452394 — Re: [PATCH] usb:serial: Add Fintek F81532/534 driver

FromOne Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk>
Date2016-07-29 14:50 +0200
SubjectRe: [PATCH] usb:serial: Add Fintek F81532/534 driver
Message-ID<s0foJ-4Gb-9@gated-at.bofh.it>
O
> +static int f81534_set_normal_register(struct usb_device *dev, u16 reg, u8 data)
> +{
> +	size_t count = F81534_USB_MAX_RETRY;
> +	int status;
> +	u8 *tmp;
> +
> +	tmp = kmalloc(sizeof(u8), GFP_KERNEL);
> +	if (!tmp)
> +		return -ENOMEM;

You end up doing huge numbers of tiny allocation and frees in some of the
code paths. I think it would be better to allocate them at a higher level
as they are not that cheap on CPU time.

> +static int f81534_read_data(struct usb_serial *usbserial, u32 address,
> +				size_t size, unsigned char *buf)
> +{

Is a particularly good example - you do 4 mallocs plus two per byte of
data.

Alan

[toc] | [next] | [standalone]


#1456187

From"Ji-Ze Hong (Peter Hong)" <hpeter@gmail.com>
Date2016-08-04 08:40 +0200
Message-ID<s2ktX-5fC-1@gated-at.bofh.it>
In reply to#1452394
Hi Alan,

One Thousand Gnomes 於 2016/7/29 下午 08:48 寫道:
> O
>> +static int f81534_set_normal_register(struct usb_device *dev, u16 reg, u8 data)
>> +{
>> +	size_t count = F81534_USB_MAX_RETRY;
>> +	int status;
>> +	u8 *tmp;
>> +
>> +	tmp = kmalloc(sizeof(u8), GFP_KERNEL);
>> +	if (!tmp)
>> +		return -ENOMEM;
>
> You end up doing huge numbers of tiny allocation and frees in some of the
> code paths. I think it would be better to allocate them at a higher level
> as they are not that cheap on CPU time.
>
>> +static int f81534_read_data(struct usb_serial *usbserial, u32 address,
>> +				size_t size, unsigned char *buf)
>> +{
>
> Is a particularly good example - you do 4 mallocs plus two per byte of
> data.
>

I'll re-factor the newest V9 patch with your suggestion. To malloc a
byte within usb_serial privates, and make a mutex to protect it.

I'll send it as V10 when I tested it.

Thanks for your suggestion.
-- 
With Best Regards,
Peter Hong

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web