[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <50fa480e-4c0e-70f2-5185-1d83df9f9e13@acm.org>
Date: Mon, 27 Mar 2023 14:25:34 -0700
From: Bart Van Assche <bvanassche@....org>
To: Ye Bin <yebin@...weicloud.com>, jejb@...ux.ibm.com,
martin.petersen@...cle.com, linux-scsi@...r.kernel.org,
linux-kernel@...r.kernel.org
Cc: Ye Bin <yebin10@...wei.com>
Subject: Re: [PATCH 3/5] scsi: limit to set the host state
On 3/24/23 18:17, Ye Bin wrote:
> From: Ye Bin <yebin10@...wei.com>
>
> Now, we can set the host state with any value. Actually, it doesn't
> make sense. As previous patch introduce SHOST_BLOCKED state, set this
> state, it will blocking IO. So this patch limit to set the host with
> running/blocked state.
>
> Signed-off-by: Ye Bin <yebin10@...wei.com>
> ---
> drivers/scsi/scsi_sysfs.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
> index b14f95ac594e..42c5936c7711 100644
> --- a/drivers/scsi/scsi_sysfs.c
> +++ b/drivers/scsi/scsi_sysfs.c
> @@ -203,6 +203,7 @@ store_shost_state(struct device *dev, struct device_attribute *attr,
> int i;
> struct Scsi_Host *shost = class_to_shost(dev);
> enum scsi_host_state state = 0;
> + enum scsi_host_state old_state;
> unsigned long flags;
>
> for (i = 0; i < ARRAY_SIZE(shost_states); i++) {
> @@ -216,8 +217,13 @@ store_shost_state(struct device *dev, struct device_attribute *attr,
> if (!state)
> return -EINVAL;
>
> + if (state != SHOST_RUNNING && state != SHOST_BLOCKED)
> + return -EINVAL;
> +
> spin_lock_irqsave(shost->host_lock, flags);
> - if (scsi_host_set_state(shost, state)) {
> + old_state = shost->shost_state;
> + if ((old_state != SHOST_RUNNING && old_state != SHOST_BLOCKED) ||
> + scsi_host_set_state(shost, state)) {
> spin_unlock_irqrestore(shost->host_lock, flags);
> return -EINVAL;
> }
Please make sure that the "state != SHOST_RUNNING && state !=
SHOST_BLOCKED" check occurs only once and also that there is one
spin_lock_irqsave() call in this function and only one
spin_unlock_irqrestore() call.
Thanks,
Bart.
Powered by blists - more mailing lists