Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20221229060549.3976570-1-ztong0001@gmail.com> <Y63WsX8Hvxep/dtN@kbusch-mbp.dhcp.thefacebook.com>
In-Reply-To: <Y63WsX8Hvxep/dtN@kbusch-mbp.dhcp.thefacebook.com>
From:   Tong Zhang <ztong0001@gmail.com>
Date:   Thu, 29 Dec 2022 10:39:15 -0800
Message-ID: <CAA5qM4Djc+5p45+isj5Vn47erUR6rY32n-2Epkkx1Pv-PX8wRg@mail.gmail.com>
Subject: Re: [PATCH] nvme-pci: free irq properly when cannot create adminq
To:     Keith Busch <kbusch@kernel.org>
Cc:     Jens Axboe <axboe@fb.com>, Christoph Hellwig <hch@lst.de>,
        Sagi Grimberg <sagi@grimberg.me>,
        Chaitanya Kulkarni <kch@nvidia.com>,
        linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
        t.zhang2@samsung.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Dec 29, 2022 at 10:04 AM Keith Busch <kbusch@kernel.org> wrote:
>
> On Wed, Dec 28, 2022 at 10:05:49PM -0800, Tong Zhang wrote:
> > nvme_pci_configure_admin_queue could return -ENODEV, in this case, we
> > will need to free IRQ properly. Otherwise following warning could be
> > triggered
> >
> > [    5.286752] WARNING: CPU: 0 PID: 33 at kernel/irq/irqdomain.c:253 irq_domain_remove+0x12d/0x140
> > [    5.290547] Call Trace:
> > [    5.290626]  <TASK>
> > [    5.290695]  msi_remove_device_irq_domain+0xc9/0xf0
> > [    5.290843]  msi_device_data_release+0x15/0x80
> > [    5.290978]  release_nodes+0x58/0x90
> > [    5.293788] WARNING: CPU: 0 PID: 33 at kernel/irq/msi.c:276 msi_device_data_release+0x76/0x80
> > [    5.297573] Call Trace:
> > [    5.297651]  <TASK>
> > [    5.297719]  release_nodes+0x58/0x90
> > [    5.297831]  devres_release_all+0xef/0x140
> > [    5.298339]  device_unbind_cleanup+0x11/0xc0
> > [    5.298479]  really_probe+0x296/0x320
> >
> > Fixes: a6ee7f19ebfd ("nvme-pci: call nvme_pci_configure_admin_queue from nvme_pci_enable")
>
> Right. It's really only needed when called from probe as the reset_work
> handles the cleanup when called from there, but this is safe for both
> cases.
>
> > @@ -2584,8 +2584,13 @@ static int nvme_pci_enable(struct nvme_dev *dev)
> >       pci_enable_pcie_error_reporting(pdev);
> >       pci_save_state(pdev);
> >
> > -     return nvme_pci_configure_admin_queue(dev);
> > +     result = nvme_pci_configure_admin_queue(dev);
> > +     if (result)
> > +             goto free_irq;
> > +     return result;
>
> Since you're already in this function, you should also add a "goto
> disable" if pci_alloc_irq_vectors() fails. Right now it just returns
> with the pci device still enabled, and it won't get disabled from probe.
>

Thank you Keith! I have added this fix and sent a v2.
- Tong

> > + free_irq:
> > +     pci_free_irq_vectors(pdev);
> >   disable:
> >       pci_disable_device(pdev);
> >       return result;
> > --