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

From: John Garry
Date: Mon Apr 25 2022 - 09:05:17 EST

On 25/04/2022 10:22, Krzysztof Kozlowski wrote:
On 25/04/2022 10:58, John Garry wrote:
On 20/04/2022 08:03, Christoph Hellwig wrote:
The standard flow is:

shost = scsi_host_alloc(sht, )

// modify shost, like
shost->cmd_per_lun = 5;


Is there some reason for which those two drivers can't follow that?
I think they should. Method tables should not be mutable data.

Hi Krzysztof,

Do you have any interest in going further with your work and trying to
change all SCSI driver instances of scsi_host_template to be const? I am
not sure if it has been attempted before...

I can work on this, but what about the SCSI core modifying the template?

I hope that this isn't a can of worms...

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)

list_for_each_entry(dir, &sht_proc_dir_list, list) {
if (dir->sht == sht) {
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);
goto out;

list_add_tail(&dir->list, &sht_proc_dir_list);

and so on..