Skip to contentSkip to conversation

check request_region return value

PATCH In review Tim Cooijmans 3 participants 12 messages Feb 24, 2006 – Feb 26, 2006 Kernel janitors

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 check request_region return value

From: Tim Cooijmans <[email protected]>

Check request_region return value in arch/powerpc/platform/chrp/setup.c. Passed compilation.

---
--- linux-2.6.16-rc4-orig/arch/powerpc/platforms/chrp/setup.c 2006-02-21
20:18:41.000000000 +0100
+++ linux-2.6.16-rc4/arch/powerpc/platforms/chrp/setup.c 2006-02-24
18:30:10.000000000 +0100
@@ -471,15 +471,54 @@ chrp_init2(void)
chrp_nvram_init();
#endif

- request_region(0x20,0x20,"pic1");
- request_region(0xa0,0x20,"pic2");
- request_region(0x00,0x20,"dma1");
- request_region(0x40,0x20,"timer");
- request_region(0x80,0x10,"dma page reg");
- request_region(0xc0,0x20,"dma2");
+ if (request_region(0x20,0x20,"pic1") = NULL) {
+ printk(KERN_WARNING "chrp_init2: could not "
+ "allocate pic1 region\n");
+ goto out_pic1;
+ }
+ if (request_region(0xa0,0x20,"pic2") = NULL) {
+ printk(KERN_WARNING "chrp_init2: could not "
+ "allocate pic2 region\n");
+ goto out_pic2;
+ }
+ if (request_region(0x00,0x20,"dma1") = NULL) {
+ printk(KERN_WARNING "chrp_init2: could not "
+ "allocate dma1 region\n");
+ goto out_dma1;
+ }
+ if (request_region(0x40,0x20,"timer") = NULL) {
+ printk(KERN_WARNING "chrp_init2: could not "
+ "allocate timer region\n");
+ goto out_timer;
+ }
+ if (request_region(0x80,0x10,"dma page reg") = NULL) {
+ printk(KERN_WARNING "chrp_init2: could not "
+ "allocate dma page reg region\n");
+ goto out_dma_page_reg;
+ }
+ if (request_region(0xc0,0x20,"dma2") = NULL) {
+ printk(KERN_WARNING "chrp_init2: could not "
+ "allocate dma2 region\n");
+ goto out_dma2;
+ }

if (ppc_md.progress)
ppc_md.progress(" Have fun! ", 0x7777);
+
+ return;
+
+out_dma2:
+ release_region(0x80,0x10);
+out_dma_page_reg:
+ release_region(0x40,0x20);
+out_timer:
+ release_region(0x00,0x20);
+out_dma1:
+ release_region(0xa0,0x20);
+out_pic2:
+ release_region(0x20,0x20);
+out_pic1:
+ return;
}

void __init chrp_init(void)
_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#1 lore raw
Quoted 5 lines · Tim Cooijmans

From: Tim Cooijmans <[email protected]>

Check request_region return value in arch/powerpc/platform/chrp/setup.c. Passed compilation.

Looks OK to me, but couldn't this be improved by changing the function to return an int stating if things are OK or not and then having caller check the return value and act appropriately?

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#2 lore raw
In reply to Tim Cooijmans
Quoted 9 lines · Jesper Juhl
Quoted 4 lines · Tim Cooijmans

From: Tim Cooijmans <[email protected]>

Check request_region return value in arch/powerpc/platform/chrp/setup.c. Passed compilation.

Looks OK to me, but couldn't this be improved by changing the function to return an int stating if things are OK or not and then having caller check the return value and act appropriately?

Yes, I thought about that. The reason I didn't do this is that chrp_init2 is assigned to ppc_md.init. ppc_md is a struct machdep_calls, which seems to be an abstraction of machine-dependent stuff. Changing the return type is probably going to break lots of things.
Anyway, warning when things fail is better than nothing.

Tim
_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#3 lore raw
In reply to Tim Cooijmans
Quoted 7 lines · Tim Cooijmans
Quoted 6 lines · Jesper Juhl
Quoted 4 lines · Tim Cooijmans

From: Tim Cooijmans <[email protected]>

Check request_region return value in arch/powerpc/platform/chrp/setup.c. Passed compilation.

Quoted 7 lines
Quoted 3 lines

Looks OK to me, but couldn't this be improved by changing the function to return an int stating if things are OK or not and then having caller check the return value and act appropriately?

Yes, I thought about that. The reason I didn't do this is that chrp_init2 is assigned to ppc_md.init. ppc_md is a struct machdep_calls, which seems to be an abstraction of machine-dependent stuff. Changing the return type is probably going to break lots of things.

Ok, I must admit I didn't look that much into it.

Quoted 2 lines

Anyway, warning when things fail is better than nothing.

Definately.

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#4 lore raw

PATCH 1/1 check request_region return value

In reply to Tim Cooijmans

From: Tim Cooijmans <[email protected]>

Check request_region return value in drivers/acpi/motherboard.c. acpi_reserve_resources now returns int, 0 being success. acpi_motherboard_init propagates that return value.

Passed compilation.

---
--- linux-2.6.16-rc4-orig/drivers/acpi/motherboard.c 2006-02-21
20:18:41.000000000 +0100
+++ linux-2.6.16-rc4/drivers/acpi/motherboard.c 2006-02-25 14:08:01.000000000
+0100
@@ -123,42 +123,115 @@ static struct acpi_driver acpi_motherboa
},
};

-static void __init acpi_reserve_resources(void)
+static int __init acpi_reserve_resources(void)
{
if (acpi_gbl_FADT->xpm1a_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)
- request_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
- acpi_gbl_FADT->pm1_evt_len, "PM1a_EVT_BLK");
+ if (request_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
+ acpi_gbl_FADT->pm1_evt_len,
+ "PM1a_EVT_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate PM1a_EVT_BLK region\n");
+ goto out_pm1a_evt_blk;
+ }

if (acpi_gbl_FADT->xpm1b_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)
- request_region(acpi_gbl_FADT->xpm1b_evt_blk.address,
- acpi_gbl_FADT->pm1_evt_len, "PM1b_EVT_BLK");
+ if (request_region(acpi_gbl_FADT->xpm1b_evt_blk.address,
+ acpi_gbl_FADT->pm1_evt_len,
+ "PM1b_EVT_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate PM1b_EVT_BLK region\n");
+ goto out_pm1b_evt_blk;
+ }

if (acpi_gbl_FADT->xpm1a_cnt_blk.address && acpi_gbl_FADT->pm1_cnt_len)
- request_region(acpi_gbl_FADT->xpm1a_cnt_blk.address,
- acpi_gbl_FADT->pm1_cnt_len, "PM1a_CNT_BLK");
+ if (request_region(acpi_gbl_FADT->xpm1a_cnt_blk.address,
+ acpi_gbl_FADT->pm1_cnt_len,
+ "PM1a_CNT_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate PM1a_CNT_BLK region\n");
+ goto out_pm1a_cnt_blk;
+ }

if (acpi_gbl_FADT->xpm1b_cnt_blk.address && acpi_gbl_FADT->pm1_cnt_len)
- request_region(acpi_gbl_FADT->xpm1b_cnt_blk.address,
- acpi_gbl_FADT->pm1_cnt_len, "PM1b_CNT_BLK");
+ if (request_region(acpi_gbl_FADT->xpm1b_cnt_blk.address,
+ acpi_gbl_FADT->pm1_cnt_len,
+ "PM1b_CNT_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate PM1b_CNT_BLK region\n");
+ goto out_pm1b_cnt_blk;
+ }

if (acpi_gbl_FADT->xpm_tmr_blk.address && acpi_gbl_FADT->pm_tm_len = 4)
- request_region(acpi_gbl_FADT->xpm_tmr_blk.address, 4, "PM_TMR");
+ if (request_region(acpi_gbl_FADT->xpm_tmr_blk.address, 4,
+ "PM_TMR") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate PM_TMR region\n");
+ goto out_pm_tmr;
+ }

if (acpi_gbl_FADT->xpm2_cnt_blk.address && acpi_gbl_FADT->pm2_cnt_len)
- request_region(acpi_gbl_FADT->xpm2_cnt_blk.address,
- acpi_gbl_FADT->pm2_cnt_len, "PM2_CNT_BLK");
+ if (request_region(acpi_gbl_FADT->xpm2_cnt_blk.address,
+ acpi_gbl_FADT->pm2_cnt_len,
+ "PM2_CNT_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate PM2_CNT_BLK region\n");
+ goto out_pm2_cnt_blk;
+ }

/* Length of GPE blocks must be a non-negative multiple of 2 */

if (acpi_gbl_FADT->xgpe0_blk.address && acpi_gbl_FADT->gpe0_blk_len &&
!(acpi_gbl_FADT->gpe0_blk_len & 0x1))
- request_region(acpi_gbl_FADT->xgpe0_blk.address,
- acpi_gbl_FADT->gpe0_blk_len, "GPE0_BLK");
+ if (request_region(acpi_gbl_FADT->xgpe0_blk.address,
+ acpi_gbl_FADT->gpe0_blk_len,
+ "GPE0_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate GPE0_BLK region\n");
+ goto out_gpe0_blk;
+ }

if (acpi_gbl_FADT->xgpe1_blk.address && acpi_gbl_FADT->gpe1_blk_len &&
!(acpi_gbl_FADT->gpe1_blk_len & 0x1))
- request_region(acpi_gbl_FADT->xgpe1_blk.address,
- acpi_gbl_FADT->gpe1_blk_len, "GPE1_BLK");
+ if (request_region(acpi_gbl_FADT->xgpe1_blk.address,
+ acpi_gbl_FADT->gpe1_blk_len,
+ "GPE1_BLK") = NULL) {
+ printk(KERN_WARNING "acpi_reserve_resources: could not"
+ "allocate GPE1_BLK region\n");
+ goto out_gpe1_blk;
+ }
+
+ return 0;
+
+out_gpe1_blk:
+ if (acpi_gbl_FADT->xgpe0_blk.address && acpi_gbl_FADT->gpe0_blk_len &&
+ !(acpi_gbl_FADT->gpe0_blk_len & 0x1))
+ release_region(acpi_gbl_FADT->xgpe0_blk.address,
+ acpi_gbl_FADT->gpe0_blk_len);
+out_gpe0_blk:
+ if (acpi_gbl_FADT->xpm2_cnt_blk.address && acpi_gbl_FADT->pm2_cnt_len)
+ release_region(acpi_gbl_FADT->xpm2_cnt_blk.address,
+ acpi_gbl_FADT->pm2_cnt_len);
+out_pm2_cnt_blk:
+ if (acpi_gbl_FADT->xpm_tmr_blk.address && acpi_gbl_FADT->pm_tm_len = 4)
+ release_region(acpi_gbl_FADT->xpm_tmr_blk.address, 4);
+out_pm_tmr:
+ if (acpi_gbl_FADT->xpm1b_cnt_blk.address && acpi_gbl_FADT->pm1_cnt_len)
+ release_region(acpi_gbl_FADT->xpm1b_cnt_blk.address,
+ acpi_gbl_FADT->pm1_cnt_len);
+out_pm1b_cnt_blk:
+ if (acpi_gbl_FADT->xpm1a_cnt_blk.address && acpi_gbl_FADT->pm1_cnt_len)
+ release_region(acpi_gbl_FADT->xpm1a_cnt_blk.address,
+ acpi_gbl_FADT->pm1_cnt_len);
+out_pm1a_cnt_blk:
+ if (acpi_gbl_FADT->xpm1b_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)
+ release_region(acpi_gbl_FADT->xpm1b_evt_blk.address,
+ acpi_gbl_FADT->pm1_evt_len);
+out_pm1b_evt_blk:
+ if (acpi_gbl_FADT->xpm1a_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)
+ release_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
+ acpi_gbl_FADT->pm1_evt_len);
+out_pm1a_evt_blk:
+ return -EBUSY;
}

static int __init acpi_motherboard_init(void)
@@ -170,7 +243,8 @@ static int __init acpi_motherboard_init(
* This module must run after scan.c
*/
if (!acpi_disabled)
- acpi_reserve_resources();
+ return acpi_reserve_resources();
+
return 0;
}

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#5 lore raw
In reply to Tim Cooijmans
Quoted 8 lines · Tim Cooijmans

From: Tim Cooijmans <[email protected]>

Check request_region return value in drivers/acpi/motherboard.c. acpi_reserve_resources now returns int, 0 being success. acpi_motherboard_init propagates that return value.

Passed compilation.

Some comments below.

Quoted 9 lines

---
--- linux-2.6.16-rc4-orig/drivers/acpi/motherboard.c 2006-02-21
20:18:41.000000000 +0100
+++ linux-2.6.16-rc4/drivers/acpi/motherboard.c 2006-02-25 14:08:01.000000000 +0100

(quoted diff) +0 -0
@@ -123,42 +123,115 @@ static struct acpi_driver acpi_motherboa
                },
 };

Review static struct acpi_driver acpi_motherboa · line 131

Quoted 10 lines
(quoted diff) +4 -3
 
-static void __init acpi_reserve_resources(void)
+static int __init acpi_reserve_resources(void)
 {
        if (acpi_gbl_FADT->xpm1a_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)
-               request_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
-                              acpi_gbl_FADT->pm1_evt_len, "PM1a_EVT_BLK");
+               if (request_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
+                                  acpi_gbl_FADT->pm1_evt_len,
+                                  "PM1a_EVT_BLK") = NULL) {

Why not

+ if (!request_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
+ acpi_gbl_FADT->pm1_evt_len,
+ "PM1a_EVT_BLK")) {

Shorter and easier to read IMHO.

Review static struct acpi_driver acpi_motherboa · line 136

Quoted 5 lines
(quoted diff) +4 -0
+                       printk(KERN_WARNING "acpi_reserve_resources: could not"
+                                           "allocate PM1a_EVT_BLK region\n");
+                       goto out_pm1a_evt_blk;
+               }
 

If you want to print a warning for each case, why not make a simple wrapper function that calls request_region and prints the warning instead of having a ton of printk()s ?

Quoted 22 lines
(quoted diff) +22 -0
+out_gpe1_blk:
+       if (acpi_gbl_FADT->xgpe0_blk.address && acpi_gbl_FADT->gpe0_blk_len &&
+           !(acpi_gbl_FADT->gpe0_blk_len & 0x1))
+               release_region(acpi_gbl_FADT->xgpe0_blk.address,
+                              acpi_gbl_FADT->gpe0_blk_len);
+out_gpe0_blk:
+       if (acpi_gbl_FADT->xpm2_cnt_blk.address && acpi_gbl_FADT->pm2_cnt_len)
+               release_region(acpi_gbl_FADT->xpm2_cnt_blk.address,
+                              acpi_gbl_FADT->pm2_cnt_len);
+out_pm2_cnt_blk:
+       if (acpi_gbl_FADT->xpm_tmr_blk.address && acpi_gbl_FADT->pm_tm_len = 4)
+               release_region(acpi_gbl_FADT->xpm_tmr_blk.address, 4);
+out_pm_tmr:
+       if (acpi_gbl_FADT->xpm1b_cnt_blk.address && acpi_gbl_FADT->pm1_cnt_len)
+               release_region(acpi_gbl_FADT->xpm1b_cnt_blk.address,
+                              acpi_gbl_FADT->pm1_cnt_len);
+out_pm1b_cnt_blk:
+       if (acpi_gbl_FADT->xpm1a_cnt_blk.address && acpi_gbl_FADT->pm1_cnt_len)
+               release_region(acpi_gbl_FADT->xpm1a_cnt_blk.address,
+                              acpi_gbl_FADT->pm1_cnt_len);
+out_pm1a_cnt_blk:
+       if (acpi_gbl_FADT->xpm1b_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)

Review static struct acpi_driver acpi_motherboa · line 168

Quoted 10 lines
(quoted diff) +8 -0
+               release_region(acpi_gbl_FADT->xpm1b_evt_blk.address,
+                              acpi_gbl_FADT->pm1_evt_len);
+out_pm1b_evt_blk:
+       if (acpi_gbl_FADT->xpm1a_evt_blk.address && acpi_gbl_FADT->pm1_evt_len)
+               release_region(acpi_gbl_FADT->xpm1a_evt_blk.address,
+                              acpi_gbl_FADT->pm1_evt_len);
+out_pm1a_evt_blk:
+       return -EBUSY;
 }
 

Why all these if's in the error path? Since you get to any of these labels in the first place you already know what request_region()s succeeded, so what's the point of checking all these variables before calling release_region()?

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#6 lore raw
In reply to Tim Cooijmans
Quoted 34 lines · Jesper Juhl
Quoted 34 lines
(quoted diff) +2 -0
+out_gpe1_blk:
+       if (acpi_gbl_FADT->xgpe0_blk.address &&

acpi_gbl_FADT->gpe0_blk_len && + !(acpi_gbl_FADT->gpe0_blk_len
& 0x1))

(quoted diff) +4 -0
+               release_region(acpi_gbl_FADT->xgpe0_blk.address,
+                              acpi_gbl_FADT->gpe0_blk_len);
+out_gpe0_blk:
+       if (acpi_gbl_FADT->xpm2_cnt_blk.address &&

acpi_gbl_FADT->pm2_cnt_len) + release_region(acpi_gbl_FADT->xpm2_cnt_blk.address, +

         acpi_gbl_FADT->pm2_cnt_len);
(quoted diff) +2 -0
+out_pm2_cnt_blk:
+       if (acpi_gbl_FADT->xpm_tmr_blk.address &&

acpi_gbl_FADT->pm_tm_len = 4) + release_region(acpi_gbl_FADT->xpm_tmr_blk.address, 4); +out_pm_tmr:
+ if (acpi_gbl_FADT->xpm1b_cnt_blk.address &&
acpi_gbl_FADT->pm1_cnt_len) + release_region(acpi_gbl_FADT->xpm1b_cnt_blk.address, +

          acpi_gbl_FADT->pm1_cnt_len);
(quoted diff) +2 -0
+out_pm1b_cnt_blk:
+       if (acpi_gbl_FADT->xpm1a_cnt_blk.address &&

acpi_gbl_FADT->pm1_cnt_len) + release_region(acpi_gbl_FADT->xpm1a_cnt_blk.address, +

          acpi_gbl_FADT->pm1_cnt_len);
(quoted diff) +2 -0
+out_pm1a_cnt_blk:
+       if (acpi_gbl_FADT->xpm1b_evt_blk.address &&

acpi_gbl_FADT->pm1_evt_len) + release_region(acpi_gbl_FADT->xpm1b_evt_blk.address, +

          acpi_gbl_FADT->pm1_evt_len);
(quoted diff) +2 -0
+out_pm1b_evt_blk:
+       if (acpi_gbl_FADT->xpm1a_evt_blk.address &&

acpi_gbl_FADT->pm1_evt_len) + release_region(acpi_gbl_FADT->xpm1a_evt_blk.address, +

          acpi_gbl_FADT->pm1_evt_len);
Quoted 8 lines

Review

Quoted 3 lines
(quoted diff) +2 -0
+out_pm1a_evt_blk:
+       return -EBUSY;
 }

Why all these if's in the error path? Since you get to any of these labels in the first place you already know what request_region()s succeeded, so what's the point of checking all these variables before calling release_region()?

Not all request_region()s are tried. These checks are done before requesting them as well, to determine which ones should be requested. If I don't do these checks here, some release_region() calls will try to release regions that were never requested.

I'll look into your other comments.

Thanks,
Tim
_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#7 lore raw
In reply to Tim Cooijmans
Quoted 43 lines · Tim Cooijmans
Quoted 34 lines · Jesper Juhl
Quoted 34 lines
(quoted diff) +2 -0
+out_gpe1_blk:
+       if (acpi_gbl_FADT->xgpe0_blk.address &&

acpi_gbl_FADT->gpe0_blk_len && + !(acpi_gbl_FADT->gpe0_blk_len
& 0x1))

(quoted diff) +4 -0
+               release_region(acpi_gbl_FADT->xgpe0_blk.address,
+                              acpi_gbl_FADT->gpe0_blk_len);
+out_gpe0_blk:
+       if (acpi_gbl_FADT->xpm2_cnt_blk.address &&

acpi_gbl_FADT->pm2_cnt_len) + release_region(acpi_gbl_FADT->xpm2_cnt_blk.address, +

         acpi_gbl_FADT->pm2_cnt_len);
(quoted diff) +2 -0
+out_pm2_cnt_blk:
+       if (acpi_gbl_FADT->xpm_tmr_blk.address &&

acpi_gbl_FADT->pm_tm_len = 4) + release_region(acpi_gbl_FADT->xpm_tmr_blk.address, 4); +out_pm_tmr:
+ if (acpi_gbl_FADT->xpm1b_cnt_blk.address &&
acpi_gbl_FADT->pm1_cnt_len) + release_region(acpi_gbl_FADT->xpm1b_cnt_blk.address, +

          acpi_gbl_FADT->pm1_cnt_len);
(quoted diff) +2 -0
+out_pm1b_cnt_blk:
+       if (acpi_gbl_FADT->xpm1a_cnt_blk.address &&

acpi_gbl_FADT->pm1_cnt_len) + release_region(acpi_gbl_FADT->xpm1a_cnt_blk.address, +

          acpi_gbl_FADT->pm1_cnt_len);
(quoted diff) +2 -0
+out_pm1a_cnt_blk:
+       if (acpi_gbl_FADT->xpm1b_evt_blk.address &&

acpi_gbl_FADT->pm1_evt_len) + release_region(acpi_gbl_FADT->xpm1b_evt_blk.address, +

          acpi_gbl_FADT->pm1_evt_len);
(quoted diff) +2 -0
+out_pm1b_evt_blk:
+       if (acpi_gbl_FADT->xpm1a_evt_blk.address &&

acpi_gbl_FADT->pm1_evt_len) + release_region(acpi_gbl_FADT->xpm1a_evt_blk.address, +

          acpi_gbl_FADT->pm1_evt_len);
Quoted 8 lines

Review

Quoted 3 lines
(quoted diff) +2 -0
+out_pm1a_evt_blk:
+       return -EBUSY;
 }

Why all these if's in the error path? Since you get to any of these labels in the first place you already know what request_region()s succeeded, so what's the point of checking all these variables before calling release_region()?

Quoted 5 lines

Not all request_region()s are tried. These checks are done before requesting them as well, to determine which ones should be requested. If I don't do these checks here, some release_region() calls will try to release regions that were never requested.

Ahh yes, looking a bit more closely at the code I see that's true. Hmm, there's got to be a cleaner way, the result of your patch is a pretty ugly error path...

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#8 lore raw
In reply to Tim Cooijmans
Quoted 3 lines · Jesper Juhl

Ahh yes, looking a bit more closely at the code I see that's true. Hmm, there's got to be a cleaner way, the result of your patch is a pretty ugly error path...

Pretty much the only alternative I can think of is something like this:

if (condition && request_region(x)) {
        if (condition && request_region(y)) {
                /* ... */
        } else {
                release_region(x);
        }
}

In this case, it will become a nine-level deep nest, and it's not much different from my original version.

Having done some research, I don't think the checks are necessary. request_region() (actually __request_resource()) does some checks which I'm pretty sure will catch any 0 passed to it. The GPE odd/even checks will still have to be done, though.

I'll make a patch without most of the checks, unless anyone objects to removing them.

Tim
_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#9 lore raw
In reply to Tim Cooijmans
Quoted 17 lines · Tim Cooijmans
Quoted 3 lines · Jesper Juhl

Ahh yes, looking a bit more closely at the code I see that's true. Hmm, there's got to be a cleaner way, the result of your patch is a pretty ugly error path...

Pretty much the only alternative I can think of is something like this:

if (condition && request_region(x)) {
        if (condition && request_region(y)) {
                /* ... */
        } else {
                release_region(x);
        }
}

In this case, it will become a nine-level deep nest, and it's not much different from my original version.

Quoted 8 lines

Having done some research, I don't think the checks are necessary. request_region() (actually __request_resource()) does some checks which I'm pretty sure will catch any 0 passed to it. The GPE odd/even checks will still have to be done, though.

I'll make a patch without most of the checks, unless anyone objects to removing them.

I just did some research myself and looked at the code currently in 2.6.16-rc4-mm2 , and the code seems to have changed quite a bit, so perhaps you would do well to work against that instead.

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#10 lore raw
In reply to Tim Cooijmans
Quoted 12 lines · Tim Cooijmans

From: Tim Cooijmans <[email protected]>

Check request_region return value in arch/powerpc/platform/chrp/setup.c. Passed compilation.

---

(quoted diff) +4 -0
+
+out_dma2:
+	release_region(0x80,0x10);
+out_dma_page_reg:

Review

Quoted 10 lines
(quoted diff) +9 -0
+	release_region(0x40,0x20);
+out_timer:
+	release_region(0x00,0x20);
+out_dma1:
+	release_region(0xa0,0x20);
+out_pic2:
+	release_region(0x20,0x20);
+out_pic1:
+	return;
 }

This might sound a bit stupid, but why are you doing release_region here ? As you can't warn the upstream code, it will presumably try to use the regions anyway. Wouldn't it be better to just to warn on a failure, and keep going as normal? leaving all the regions (that were acquired) mapped, keeping breakage/(usage of unmapped regions) to a minimum?

Darren Jenkins

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors

#11 lore raw
In reply to Tim Cooijmans
Quoted 6 lines · Darren Jenkins\

This might sound a bit stupid, but why are you doing release_region here ? As you can't warn the upstream code, it will presumably try to use the regions anyway. Wouldn't it be better to just to warn on a failure, and keep going as normal? leaving all the regions (that were acquired) mapped, keeping breakage/(usage of unmapped regions) to a minimum?

Good point. I'll revise my patch and resubmit.

Tim

_______________________________________________
Kernel-janitors mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/kernel-janitors