Re: [PATCH 1/4] scsi: core: constify pointer to scsi_host_template

[Douglas Gilbert]

On 2022-04-25 21:16, Bart Van Assche wrote:
> On 4/25/22 06:04, John Garry wrote:
> > On 25/04/2022 10:22, Krzysztof Kozlowski wrote:
> > > For example scsi_proc_hostdir_rm(): 'present' and 'proc_dir' members.
> > > Where should they be stored? Should they be moved to the Scsi_Host?
> >
> > I don't think scsi_Host is appropriate as this is per-scsi host template, 
> > unless you see a way to do it that way. Alternatively we could keep a separate 
> > list of registered sht, like this:
> >
> > struct sht_proc_dir {
> >      int cnt;
> >      struct list_head list;
> >      struct proc_dir_entry *proc_dir;
> >      struct scsi_host_template *sht;
> > };
> > static LIST_HEAD(sht_proc_dir_list);
> >
> > void scsi_proc_hostdir_add(struct scsi_host_template *sht)
> > {
> >      struct sht_proc_dir *dir;
> >
> >      if (!sht->show_info)
> >          return;
> >
> >      mutex_lock(&global_host_template_mutex);
> >      list_for_each_entry(dir, &sht_proc_dir_list, list) {
> >          if (dir->sht == sht) {
> >              dir->cnt++;
> >              goto out;
> >          }
> >      }
> >      dir = kzalloc(sizeof(*dir), GFP_KERNEL);
> >      if (!dir)
> >          goto out;
> >
> >      dir->proc_dir = proc_mkdir(sht->proc_name, proc_scsi);
> >      if (!dir->proc_dir) {
> >          printk(KERN_ERR "%s: proc_mkdir failed for %s\n",
> >                     __func__, sht->proc_name);
> >          kfree(dir);
> >          goto out;
> >      }
> >
> >      dir->cnt++;
> >      list_add_tail(&dir->list, &sht_proc_dir_list);
> > out:
> >      mutex_unlock(&global_host_template_mutex);
> > }
>
> How about removing scsi_proc_hostdir_add(), scsi_proc_hostdir_rm() and all other 
> code that creates files or directories under /proc/scsi? There should be 
> corresponding entries in sysfs for all /proc/scsi entries. Some tools in 
> sg3_utils use that directory so sg3_utils will have to be updated.

... breaking this:

~$ cat /proc/scsi/scsi

Attached devices:

Host: scsi3 Channel: 00 Id: 00 Lun: 00

  Vendor: IBM-207x Model: HUSMM8020ASS20   Rev: J4B6

  Type:   Direct-Access                    ANSI  SCSI revision: 06

Host: scsi3 Channel: 00 Id: 01 Lun: 00

  Vendor: IBM-207x Model: HUSMM8020ASS20   Rev: J4B6

  Type:   Direct-Access                    ANSI  SCSI revision: 06

Host: scsi3 Channel: 00 Id: 02 Lun: 00

  Vendor: SEAGATE  Model: ST200FM0073      Rev: 0007

  Type:   Direct-Access                    ANSI  SCSI revision: 06
...

A deprecation notice would be helpful, then removal after a few kernel
cycles. Yes, lsscsi can give that output:

$ lsscsi -c

Attached devices:

Host: scsi2 Channel: 00 Target: 00 Lun: 00

  Vendor: SEAGATE  Model: ST200FM0073      Rev: 0007

  Type:   Direct-Access                    ANSI SCSI revision: 06

Host: scsi2 Channel: 00 Target: 01 Lun: 00

  Vendor: WDC      Model: WSH722020AL5204  Rev: C421

  Type:   Zoned Block                      ANSI SCSI revision: 07

Host: scsi2 Channel: 00 Target: 02 Lun: 00

  Vendor: Areca Te Model: ARC-802801.37.69 Rev: 0137

  Type:   Enclosure                        ANSI SCSI revision: 05
...

[Hmmm, in a different order.]

However no distribution that I'm aware of includes lsscsi in its installation.
[Most recent example: Ubuntu 22.04]
Linux is not alone ... in FreeBSD how do you list SCSI devices in your
system? Answer: as root you invoke 'camcontrol devlist', it's so obvious.

Perhaps the Linux kernel could have a deprecation process which uses inotify
or similar to notice accesses to /proc/scsi/scsi (say) and print out
a helpful response along the lines" "this is no longer supported, try using
the lsscsi utility".

Doug Gilbert