linux-kernel - Re: [PATCH v3 4/4] nvme-fc: wait for initial connect attempt to finish

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <58813044-1e6f-4267-bd10-70cc96348ecf@flourine.local>
Date: Wed, 3 Sep 2025 09:45:38 +0200
From: Daniel Wagner <dwagner@...e.de>
To: Hannes Reinecke <hare@...e.de>
Cc: Daniel Wagner <wagi@...nel.org>, Keith Busch <kbusch@...nel.org>, 
	Christoph Hellwig <hch@....de>, Sagi Grimberg <sagi@...mberg.me>, 
	James Smart <james.smart@...adcom.com>, Shinichiro Kawasaki <shinichiro.kawasaki@....com>, 
	linux-nvme@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 4/4] nvme-fc: wait for initial connect attempt to
 finish

On Tue, Sep 02, 2025 at 11:13:41AM +0200, Hannes Reinecke wrote:
> > +	if (!opts->connect_async) {
> > +		enum nvme_ctrl_state state;
> > +
> > +		wait_for_completion(&ctrl->connect_completion);
> > +		state = nvme_ctrl_state(&ctrl->ctrl);
> > +		nvme_fc_ctrl_put(ctrl);
> > +
> > +		if (state != NVME_CTRL_LIVE) {
> > +			/* Cleanup is handled by the connect state machine */
> > +			pr_info("%s:%d ctrl state %d\n", __func__, __LINE__, state);
> > +			return ERR_PTR(-EIO);
> 
> We really should return the correct status (and not just -EIO).
> Guess we'll need to introduce another variable in struct nvme_fc_ctrl
> to hold the connect status...

(forgot to remove the debug prints)

But this is what the current return value is if something goes wrong.
The other transport map all status codes to -EIO or return the negative
error code from the function which fails. I'd rather not change this in
this series.

> >   	dev_info(ctrl->ctrl.device,
> >   		"NVME-FC{%d}: new ctrl: NQN \"%s\", hostnqn: %s\n",
> >   		ctrl->cnum, nvmf_ctrl_subsysnqn(&ctrl->ctrl), opts->host->nqn);
> > @@ -3895,6 +3928,7 @@ nvme_fc_delete_controllers(struct nvme_fc_rport *rport)
> >   		dev_warn(ctrl->ctrl.device,
> >   			"NVME-FC{%d}: transport unloading: deleting ctrl\n",
> >   			ctrl->cnum);
> > +		complete(&ctrl->connect_completion);
> >   		nvme_fc_ctrl_put(ctrl);
> >   	}
> >   	spin_unlock(&rport->lock);
> > 
> 
> And I wonder: what about the udev rules?
> Do they need to be modified?
> (IE: should we call udev with --connect-async or without?)

Without changing user space, the old behavior will be used. It's an
opt-in feature. I haven't finished the libnvme/nvme-cli changes but I
expect for getting all working the udev rules need to be modified as
well. My plan is to get the kernel bits finished up first before
tackling all the user bits in detail, e.g. also creating blktests for
this (in a way we have the first one which is nvme/041)

I'll post those changes here as it needs wider review for sure.