DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Zhang, Qi Z" <qi.z.zhang@intel.com>
To: Thomas Monjalon <thomas@monjalon.net>, "dev@dpdk.org" <dev@dpdk.org>
Cc: "stable@dpdk.org" <stable@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 3/3] eal: fix multi-process probe failure handling
Date: Thu, 14 Mar 2019 07:15:53 +0000	[thread overview]
Message-ID: <039ED4275CED7440929022BC67E706115334C7D0@SHSMSX103.ccr.corp.intel.com> (raw)
Message-ID: <20190314071553.s0mCn2c1T_RsCM4WoIFftuxirPzu3UnmyAmKNbAb0no@z> (raw)
In-Reply-To: <20190302024253.15594-4-thomas@monjalon.net>



> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas@monjalon.net]
> Sent: Saturday, March 2, 2019 10:43 AM
> To: dev@dpdk.org
> Cc: Zhang, Qi Z <qi.z.zhang@intel.com>; stable@dpdk.org
> Subject: [PATCH 3/3] eal: fix multi-process probe failure handling
> 
> If probe fails in multi-process context, the device must removed in other
> processes for consistency. This is a rollback mechanism.
> However the rollback should not happen for devices which were already probed
> before the current probe transaction.
> 
> When probing an already probed device, the driver may reject with -EEXIST or
> update and succeed with code 0.
> In order to distinguish successful new probe from re-probe, in the function
> local_dev_probe(), the positive EEXIST code is returned for the latter case.
> 
> The functions rte_dev_probe() and __handle_secondary_request() can test for
> -EEXIST and +EEXIST, and skip rollback in such case.
> 
> Fixes: 244d5130719c ("eal: enable hotplug on multi-process")
> Fixes: ac9e4a17370f ("eal: support attach/detach shared device from secondary")
> Cc: qi.z.zhang@intel.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> ---
>  lib/librte_eal/common/eal_common_dev.c | 12 ++++++++++--
>  lib/librte_eal/common/eal_private.h    |  2 +-
>  lib/librte_eal/common/hotplug_mp.c     |  8 ++++++--
>  3 files changed, 17 insertions(+), 5 deletions(-)
> 
> diff --git a/lib/librte_eal/common/eal_common_dev.c
> b/lib/librte_eal/common/eal_common_dev.c
> index deaaea9345..2c7b1ab071 100644
> --- a/lib/librte_eal/common/eal_common_dev.c
> +++ b/lib/librte_eal/common/eal_common_dev.c
> @@ -132,6 +132,7 @@ local_dev_probe(const char *devargs, struct rte_device
> **new_dev)  {
>  	struct rte_device *dev;
>  	struct rte_devargs *da;
> +	bool already_probed;
>  	int ret;
> 
>  	*new_dev = NULL;
> @@ -171,12 +172,15 @@ local_dev_probe(const char *devargs, struct rte_device
> **new_dev)
>  	 * those devargs shouldn't be removed manually anymore.
>  	 */
> 
> +	already_probed = rte_dev_is_probed(dev);
>  	ret = dev->bus->plug(dev);
>  	if (ret && !rte_dev_is_probed(dev)) { /* if hasn't ever succeeded */
>  		RTE_LOG(ERR, EAL, "Driver cannot attach the device (%s)\n",
>  			dev->name);
>  		return ret;
>  	}
> +	if (ret == 0 && already_probed)
> +		ret = EEXIST; /* hint to avoid any rollback */

What if bus->plug return -EEXIST and rte_dev_is_probed return true? (See rte_pci_probe_one_driver)
You will not give hint here, but is this expected?

> 
>  	*new_dev = dev;
>  	return ret;
> @@ -194,6 +198,7 @@ rte_dev_probe(const char *devargs)  {
>  	struct eal_dev_mp_req req;
>  	struct rte_device *dev;
> +	bool already_probed;
>  	int ret;
> 
>  	memset(&req, 0, sizeof(req));
> @@ -221,8 +226,8 @@ rte_dev_probe(const char *devargs)
> 
>  	/* primary attach the new device itself. */
>  	ret = local_dev_probe(devargs, &dev);
> -
> -	if (ret != 0 && ret != -EEXIST) {
> +	already_probed = (ret == -EEXIST || ret == EEXIST);
> +	if (ret < 0 && !already_probed) {
>  		RTE_LOG(ERR, EAL,
>  			"Failed to attach device on primary process\n");
>  		return ret;
> @@ -250,6 +255,9 @@ rte_dev_probe(const char *devargs)
>  	return 0;
> 
>  rollback:
> +	if (already_probed)
> +		return ret; /* skip rollback */
> +
>  	req.t = EAL_DEV_REQ_TYPE_ATTACH_ROLLBACK;
> 
>  	/* primary send rollback request to secondary. */ diff --git
> a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
> index 798ede553b..a01d252930 100644
> --- a/lib/librte_eal/common/eal_private.h
> +++ b/lib/librte_eal/common/eal_private.h
> @@ -304,7 +304,7 @@ rte_devargs_layers_parse(struct rte_devargs *devargs,
>   * @param new_dev
>   *   new device be probed as output.
>   * @return
> - *   0 on success, negative on error.
> + *   >=0 on success (+EEXIST if already probed), negative on error.
>   */
>  int local_dev_probe(const char *devargs, struct rte_device **new_dev);
> 
> diff --git a/lib/librte_eal/common/hotplug_mp.c
> b/lib/librte_eal/common/hotplug_mp.c
> index 69e9a16d6a..9f8ef28a3b 100644
> --- a/lib/librte_eal/common/hotplug_mp.c
> +++ b/lib/librte_eal/common/hotplug_mp.c
> @@ -90,13 +90,15 @@ __handle_secondary_request(void *param)
>  	struct rte_devargs da;
>  	struct rte_device *dev;
>  	struct rte_bus *bus;
> +	bool already_probed = false;
>  	int ret = 0;
> 
>  	tmp_req = *req;
> 
>  	if (req->t == EAL_DEV_REQ_TYPE_ATTACH) {
>  		ret = local_dev_probe(req->devargs, &dev);
> -		if (ret != 0 && ret != -EEXIST) {
> +		already_probed = (ret == -EEXIST || ret == EEXIST);
> +		if (ret < 0 && !already_probed) {
>  			RTE_LOG(ERR, EAL, "Failed to hotplug add device on primary\n");
>  			goto finish;
>  		}
> @@ -159,7 +161,7 @@ __handle_secondary_request(void *param)
>  	goto finish;
> 
>  rollback:
> -	if (req->t == EAL_DEV_REQ_TYPE_ATTACH) {
> +	if (req->t == EAL_DEV_REQ_TYPE_ATTACH && !already_probed) {
>  		tmp_req.t = EAL_DEV_REQ_TYPE_ATTACH_ROLLBACK;
>  		eal_dev_hotplug_request_to_secondary(&tmp_req);
>  		local_dev_remove(dev);
> @@ -238,6 +240,8 @@ static void __handle_primary_request(void *param)
>  	case EAL_DEV_REQ_TYPE_ATTACH:
>  	case EAL_DEV_REQ_TYPE_DETACH_ROLLBACK:
>  		ret = local_dev_probe(req->devargs, &dev);
> +		if (ret > 0)
> +			ret = 0; /* return only errors */
>  		break;
>  	case EAL_DEV_REQ_TYPE_DETACH:
>  	case EAL_DEV_REQ_TYPE_ATTACH_ROLLBACK:
> --
> 2.20.1


  reply	other threads:[~2019-03-14  7:17 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-03-02  2:42 [dpdk-dev] [PATCH 0/3] fix error path of multi-process probe Thomas Monjalon
2019-03-02  2:42 ` [dpdk-dev] [PATCH 1/3] eal: remove useless checks for already probed device Thomas Monjalon
2019-03-13 13:46   ` Zhang, Qi Z
2023-06-11  2:52     ` Stephen Hemminger
2019-03-02  2:42 ` [dpdk-dev] [PATCH 2/3] eal: remove error logs " Thomas Monjalon
2019-03-02  2:42 ` [dpdk-dev] [PATCH 3/3] eal: fix multi-process probe failure handling Thomas Monjalon
2019-03-14  7:15   ` Zhang, Qi Z [this message]
2019-03-14  7:15     ` Zhang, Qi Z
2023-06-14 19:39 ` [dpdk-dev] [PATCH 0/3] fix error path of multi-process probe Stephen Hemminger

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=039ED4275CED7440929022BC67E706115334C7D0@SHSMSX103.ccr.corp.intel.com \
    --to=qi.z.zhang@intel.com \
    --cc=dev@dpdk.org \
    --cc=stable@dpdk.org \
    --cc=thomas@monjalon.net \
    /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).