2022-06-07 15:35:31

by Jiri Slaby

Subject: [PATCH 33/36] tty/vt: consolemap: change refcount only if needed in con_do_clear_unimap()

con_do_clear_unimap() currently decreases and increases refcount of old
dictionary in a back and forth fashion. This makes the code really hard
to follow. Decrease the refcount only if everything went well and we
really allocated a new one and decoupled from the old dictionary.

I sincerelly hope I did not make a mistake in this (ill) logic.

Signed-off-by: Jiri Slaby <[email protected]>
---
drivers/tty/vt/consolemap.c | 13 +++++++------
1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/drivers/tty/vt/consolemap.c b/drivers/tty/vt/consolemap.c
index 01b7e49f1f91..4d8efe74315c 100644
--- a/drivers/tty/vt/consolemap.c
+++ b/drivers/tty/vt/consolemap.c
@@ -535,22 +535,23 @@ static int con_do_clear_unimap(struct vc_data *vc)
{
struct uni_pagedict *old = *vc->vc_uni_pagedir_loc;

- if (!old || --old->refcount) {
+ if (!old || old->refcount > 1) {
struct uni_pagedict *new = kzalloc(sizeof(*new), GFP_KERNEL);
- if (!new) {
- if (old)
- old->refcount++;
+ if (!new)
return -ENOMEM;
- }
+
new->refcount = 1;
*vc->vc_uni_pagedir_loc = new;
+
+ if (old)
+ old->refcount--;
} else {
if (old == dflt)
dflt = NULL;
- old->refcount++;
old->sum = 0;
con_release_unimap(old);
}
+
return 0;
}

--
2.36.1

2022-06-07 18:12:35

by Ilpo Järvinen

Subject: Re: [PATCH 33/36] tty/vt: consolemap: change refcount only if needed in con_do_clear_unimap()

On Tue, 7 Jun 2022, Jiri Slaby wrote:

> con_do_clear_unimap() currently decreases and increases refcount of old
> dictionary in a back and forth fashion. This makes the code really hard
> to follow. Decrease the refcount only if everything went well and we
> really allocated a new one and decoupled from the old dictionary.
>
> I sincerelly hope I did not make a mistake in this (ill) logic.
>
> Signed-off-by: Jiri Slaby <[email protected]>

It seems fine:

Reviewed-by: Ilpo J?rvinen <[email protected]>

One unrelated comment below for additional cleanup opportunity.

> ---
> drivers/tty/vt/consolemap.c | 13 +++++++------
> 1 file changed, 7 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/tty/vt/consolemap.c b/drivers/tty/vt/consolemap.c
> index 01b7e49f1f91..4d8efe74315c 100644
> --- a/drivers/tty/vt/consolemap.c
> +++ b/drivers/tty/vt/consolemap.c
> @@ -535,22 +535,23 @@ static int con_do_clear_unimap(struct vc_data *vc)
> {
> struct uni_pagedict *old = *vc->vc_uni_pagedir_loc;
>
> - if (!old || --old->refcount) {
> + if (!old || old->refcount > 1) {
> struct uni_pagedict *new = kzalloc(sizeof(*new), GFP_KERNEL);
> - if (!new) {
> - if (old)
> - old->refcount++;
> + if (!new)
> return -ENOMEM;
> - }
> +
> new->refcount = 1;
> *vc->vc_uni_pagedir_loc = new;
> +
> + if (old)
> + old->refcount--;
> } else {
> if (old == dflt)
> dflt = NULL;

This seems unnecessary as con_release_unimap() already does it.

> - old->refcount++;
> old->sum = 0;
> con_release_unimap(old);
> }
> +
> return 0;
> }
>
>

--
i.

2022-06-08 08:57:21

by Jiri Slaby

Subject: Re: [PATCH 33/36] tty/vt: consolemap: change refcount only if needed in con_do_clear_unimap()

On 07. 06. 22, 17:31, Ilpo Järvinen wrote:
> On Tue, 7 Jun 2022, Jiri Slaby wrote:
>
>> con_do_clear_unimap() currently decreases and increases refcount of old
>> dictionary in a back and forth fashion. This makes the code really hard
>> to follow. Decrease the refcount only if everything went well and we
>> really allocated a new one and decoupled from the old dictionary.
>>
>> I sincerelly hope I did not make a mistake in this (ill) logic.
>>
>> Signed-off-by: Jiri Slaby <[email protected]>
>
> It seems fine:
>
> Reviewed-by: Ilpo Järvinen <[email protected]>
>
> One unrelated comment below for additional cleanup opportunity.
>
>> ---
>> drivers/tty/vt/consolemap.c | 13 +++++++------
>> 1 file changed, 7 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/tty/vt/consolemap.c b/drivers/tty/vt/consolemap.c
>> index 01b7e49f1f91..4d8efe74315c 100644
>> --- a/drivers/tty/vt/consolemap.c
>> +++ b/drivers/tty/vt/consolemap.c
>> @@ -535,22 +535,23 @@ static int con_do_clear_unimap(struct vc_data *vc)
>> {
>> struct uni_pagedict *old = *vc->vc_uni_pagedir_loc;
>>
>> - if (!old || --old->refcount) {
>> + if (!old || old->refcount > 1) {
>> struct uni_pagedict *new = kzalloc(sizeof(*new), GFP_KERNEL);
>> - if (!new) {
>> - if (old)
>> - old->refcount++;
>> + if (!new)
>> return -ENOMEM;
>> - }
>> +
>> new->refcount = 1;
>> *vc->vc_uni_pagedir_loc = new;
>> +
>> + if (old)
>> + old->refcount--;
>> } else {
>> if (old == dflt)
>> dflt = NULL;
>
> This seems unnecessary as con_release_unimap() already does it.

Good point -- the code is real mess... Now, the mess is more apparent,
at least ;).

I will create a separate patch for this.

thanks,
--
js
suse labs