DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Hunt, David" <david.hunt@intel.com>
To: Radoslaw Biernacki <radoslaw.biernacki@linaro.org>, dev@dpdk.org
Cc: stable@dpdk.org, alan.carew@intel.com, pablo.de.lara.guarch@intel.com
Subject: Re: [dpdk-dev] [PATCH 1/2] power: switching to unbuffered stdio for /sys file access
Date: Wed, 18 Oct 2017 11:28:12 +0100	[thread overview]
Message-ID: <5ed2fa68-31b6-41aa-8589-b6f803dd9898@intel.com> (raw)
In-Reply-To: <1508161628-4265-1-git-send-email-radoslaw.biernacki@linaro.org>

Hi Radoslaw,

Thanks for the patch. Some comments below.


On 16/10/2017 2:47 PM, Radoslaw Biernacki wrote:
> This patch fixes the bug caused by improper use of buffered
> stdio file access for switching the CPU frequency and
> governor. Each write operation when using buffered stdio
> must be flushed out and the return code from fflush() must
> be verified. In buffered mode, write() syscall return value
> is is not returned by fwrite()/fputs()/fprintf().
> Since with buffered approatch, fflush() need to be done
> every time it is better to use unbuffered mode or not use
> stdio at all (instead use plain open/write functions). To
> minimize amount of changes this fix use first approach.
>
> Signed-off-by: Radoslaw Biernacki <radoslaw.biernacki@linaro.org>
> ---
>   lib/librte_power/rte_power_acpi_cpufreq.c | 18 +++++++++++++++++-
>   1 file changed, 17 insertions(+), 1 deletion(-)
>
> diff --git a/lib/librte_power/rte_power_acpi_cpufreq.c b/lib/librte_power/rte_power_acpi_cpufreq.c
> index 01ac5ac..8bf5685 100644
> --- a/lib/librte_power/rte_power_acpi_cpufreq.c
> +++ b/lib/librte_power/rte_power_acpi_cpufreq.c
> @@ -143,12 +143,13 @@ set_freq_internal(struct rte_power_info *pi, uint32_t idx)
>   				"for setting frequency for lcore %u\n", pi->lcore_id);
>   		return -1;
>   	}
> +	/* we use unbuffered mode so following will fail if kernel will refuse
> +	 * freq setting */
>   	if (fprintf(pi->f, "%u", pi->freqs[idx]) < 0) {
>   		RTE_LOG(ERR, POWER, "Fail to write new frequency for "
>   				"lcore %u\n", pi->lcore_id);
>   		return -1;
>   	}
> -	fflush(pi->f);
>   	pi->curr_idx = idx;
>   

One suggestion to to avoid switching to unbuffered mode would be to move 
the check from the fprintf to the fflush. That would eliminate the need 
for the changes below, keeping the changes to a minimum.
  But I also agree that open/write would be a better option, should you 
prefer to go that route.

>   	return 1;
> @@ -174,6 +175,11 @@ power_set_governor_userspace(struct rte_power_info *pi)
>   	f = fopen(fullpath, "rw+");
>   	FOPEN_OR_ERR_RET(f, ret);
>   
> +	if (setvbuf(f, NULL, _IONBF, 0)) {
> +		RTE_LOG(ERR, POWER, "Cannot set unbuffered mode\n");
> +		goto out;
> +	}
> +
>   	s = fgets(buf, sizeof(buf), f);
>   	FOPS_OR_NULL_GOTO(s, out);
>   
> @@ -292,6 +298,11 @@ power_init_for_setting_freq(struct rte_power_info *pi)
>   	f = fopen(fullpath, "rw+");
>   	FOPEN_OR_ERR_RET(f, -1);
>   
> +	if (setvbuf(f, NULL, _IONBF, 0)) {
> +		RTE_LOG(ERR, POWER, "Cannot set unbuffered mode\n");
> +		goto out;
> +	}
> +
>   	s = fgets(buf, sizeof(buf), f);
>   	FOPS_OR_NULL_GOTO(s, out);
>   
> @@ -389,6 +400,11 @@ power_set_governor_original(struct rte_power_info *pi)
>   	f = fopen(fullpath, "rw+");
>   	FOPEN_OR_ERR_RET(f, ret);
>   
> +	if (setvbuf(f, NULL, _IONBF, 0)) {
> +		RTE_LOG(ERR, POWER, "Cannot set unbuffered mode\n");
> +		goto out;
> +	}
> +
>   	s = fgets(buf, sizeof(buf), f);
>   	FOPS_OR_NULL_GOTO(s, out);
>   

Regards,
Dave

  parent reply	other threads:[~2017-10-18 10:28 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-10-16 13:47 Radoslaw Biernacki
2017-10-16 13:47 ` [dpdk-dev] [PATCH 2/2] power: check if userspace governor is available Radoslaw Biernacki
2017-10-18 10:53   ` Hunt, David
2017-10-18 14:02     ` Radoslaw Biernacki
2017-10-18 10:28 ` Hunt, David [this message]
2017-10-18 10:33 ` [dpdk-dev] [PATCH 1/2] power: switching to unbuffered stdio for /sys file access Hunt, David
2017-10-18 13:56   ` Radoslaw Biernacki
2017-11-11 18:55 ` [dpdk-dev] [PATCH 0/3] power: fixes for power ACPI through sysfs Radoslaw Biernacki
2017-11-11 18:55   ` [dpdk-dev] [PATCH 1/3] power: removing code macros Radoslaw Biernacki
2017-11-11 18:55   ` [dpdk-dev] [PATCH 2/3] power: switching to unbuffered access for /sys files Radoslaw Biernacki
2017-11-23 14:42     ` Hunt, David
2017-12-01 21:01       ` Radoslaw Biernacki
2018-01-31 23:38         ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
2017-11-11 18:55   ` [dpdk-dev] [PATCH 3/3] power: check if userspace governor is available Radoslaw Biernacki
2017-11-21 11:09   ` [dpdk-dev] [PATCH 0/3] power: fixes for power ACPI through sysfs Radoslaw Biernacki

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=5ed2fa68-31b6-41aa-8589-b6f803dd9898@intel.com \
    --to=david.hunt@intel.com \
    --cc=alan.carew@intel.com \
    --cc=dev@dpdk.org \
    --cc=pablo.de.lara.guarch@intel.com \
    --cc=radoslaw.biernacki@linaro.org \
    --cc=stable@dpdk.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).