OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.
Subsystems (4)
PATCH 1/1 OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.
Some clock nodes like wdt1_fck, sr1_fck, sr2_fck etc do not have a clock domain associated . For such nodes accessing the clock domain pointers in _add_initiator_dep and _del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.
---
This patch depends on http://patchwork.kernel.org/patch/63383/
arch/arm/mach-omap2/omap_hwmod.c | 8 ++++++-- 1 files changed, 6 insertions(+), 2 deletions(-)
arch/arm/mach-omap2/omap_hwmod.c +6 -2
@@ -314,8 +314,10 @@ static int _add_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod *init_oh)if (!oh->_clk)return -EINVAL ; -return pwrdm_add_sleepdep (oh->_clk->clkdm->pwrdm.ptr, +if (oh->_clk->clkdm) +return pwrdm_add_sleepdep (oh->_clk->clkdm->pwrdm.ptr, init_oh->_clk->clkdm->pwrdm.ptr); +return 0 ; }/**
@@ -335,8 +337,10 @@ static int _del_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod *init_oh)if (!oh->_clk)return -EINVAL ; -return pwrdm_del_sleepdep (oh->_clk->clkdm->pwrdm.ptr, +if (oh->_clk->clkdm) +return pwrdm_del_sleepdep (oh->_clk->clkdm->pwrdm.ptr, init_oh->_clk->clkdm->pwrdm.ptr); +return 0 ; }/**
Signature · 1 line
1.5.4.7
Quoted 8 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.
Quoted 4 lines
Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Kevin
Quoted 38 lines
arch/arm/mach-omap2/omap_hwmod.c | 8 ++++++-- 1 files changed, 6 insertions(+), 2 deletions(-)arch/arm/mach-omap2/omap_hwmod.c +6
-2@@ -314,8 +314,10 @@ static int _add_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod *init_oh)if (!oh->_clk)return -EINVAL ; -return pwrdm_add_sleepdep (oh->_clk->clkdm->pwrdm.ptr, +if (oh->_clk->clkdm) +return pwrdm_add_sleepdep (oh->_clk->clkdm->pwrdm.ptr, init_oh->_clk->clkdm->pwrdm.ptr); +return 0 ; }/** @@ -335,8 +337,10 @@ static int _del_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod *init_oh)if (!oh->_clk)return -EINVAL ; -return pwrdm_del_sleepdep (oh->_clk->clkdm->pwrdm.ptr, +if (oh->_clk->clkdm) +return pwrdm_del_sleepdep (oh->_clk->clkdm->pwrdm.ptr, init_oh->_clk->clkdm->pwrdm.ptr); +return 0 ; }/** Signature · 6 lines
1.5.4.7 -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 25 lines
Quoted 25 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 4:39 AM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.Quoted 12 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/
Quoted 5 lines
Quoted 5 lines
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Are you suggesting adding a flag to the clock node in case it does not have an associated clockdomain?
I am using your tree origin/pm-wip/omap_device branch. On this I see a commit (e89a95db7095d998991c564a756a33ee0ec722c5) in which a warning is printed in omap2_init_clk_clkdm if there is no clockdomain associated with a clock node. My idea was this warning plus the check so as to avoid the system crash would suffice.
Quoted 44 lines
Quoted 44 lines
Kevin
arch/arm/mach-omap2/omap_hwmod.c static int _add_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod
Quoted 8 lines
arch/arm/mach-omap2/omap_hwmod.c | 8 ++++++-- 1 files changed, 6 insertions(+), 2 deletions(-)arch/arm/mach-omap2/omap_hwmod.c +0
-0@@ -314,8 +314,10 @@ static int _add_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod
*init_oh)arch/arm/mach-omap2/omap_hwmod.c static int _del_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod
Quoted 12 lines
arch/arm/mach-omap2/omap_hwmod.c +3
-1if (!oh->_clk)return -EINVAL ; -return pwrdm_add_sleepdep (oh->_clk->clkdm->pwrdm.ptr, +if (oh->_clk->clkdm) +return pwrdm_add_sleepdep (oh->_clk->clkdm->pwrdm.ptr, init_oh->_clk->clkdm->pwrdm.ptr); +return 0 ; }/** @@ -335,8 +337,10 @@ static int _del_initiator_dep(struct omap_hwmod *oh, struct omap_hwmod
*init_oh)Quoted 18 lines
arch/arm/mach-omap2/omap_hwmod.c +3
-2if (!oh->_clk)return -EINVAL ; -return pwrdm_del_sleepdep (oh->_clk->clkdm->pwrdm.ptr, +if (oh->_clk->clkdm) +return pwrdm_del_sleepdep (oh->_clk->clkdm->pwrdm.ptr, init_oh->_clk->clkdm->pwrdm.ptr); +return 0 ; }/** -- 1.5.4.7
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 23 lines · Thara
Quoted 23 lines
Quoted 23 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 4:39 AM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.Quoted 12 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Are you suggesting adding a flag to the clock node in case it does not have an associated clockdomain?
Yes, or to the hwmod.
Quoted 4 lines
I am using your tree origin/pm-wip/omap_device branch. On this I see a commit (
e89a95db7095d998991c564a756a33ee0ec722c5) in which a warning is printed inomap2_init_clk_clkdmif there is no clockdomain associated with a clock node.
This is to check for the case of a missing clkdm when there should be one. You're trying to solve the problem of when there is no clockdomain for a given clock.
Quoted 2 lines
My idea was this warning plus the check so as to avoid the system crash would suffice.
I still think this should be flagged explicitly in the clock/hwmod, then check using the flag instead of of using NULL pointer check.
For readability, it makes it more clear that this is checking for some exceptions and not simply doing error checking.
Kevin
Quoted 56 lines
Quoted 56 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 10:36 PM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.Quoted 6 lines · Thara
Quoted 6 lines
Quoted 6 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 4:39 AM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change thesleep
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
dependencies.
Quoted 12 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Are you suggesting adding a flag to the clock node in case it does not have an associated clockdomain?
Yes, or to the hwmod.
Quoted 4 lines
I am using your tree origin/pm-wip/omap_device branch. On this I see a commit (
e89a95db7095d998991c564a756a33ee0ec722c5) in which a warning is printed inomap2_init_clk_clkdmif there is no clockdomain associated with a clock node.This is to check for the case of a missing clkdm when there should be one. You're trying to solve the problem of when there is no clockdomain for a given clock.
Quoted 8 lines
Quoted 8 lines
Quoted 2 lines
My idea was this warning plus the check so as to avoid the system crash would suffice.
I still think this should be flagged explicitly in the clock/hwmod, then check using the flag instead of of using NULL pointer check.
For readability, it makes it more clear that this is checking for some exceptions and not simply doing error checking.
Agreed. We could add a flag and base the check on the flag. But still if some clock node does not have a clkdomain and forgets to put this flag, the system will still crash. So have have both the checks ??
Quoted 2 lines
Quoted 2 lines
Kevin
Quoted 60 lines · Thara
Quoted 60 lines
Quoted 60 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 10:36 PM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.Quoted 6 lines · Thara
Quoted 6 lines
Quoted 6 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 4:39 AM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change thesleep
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
dependencies.
Quoted 12 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Are you suggesting adding a flag to the clock node in case it does not have an associated clockdomain?
Yes, or to the hwmod.
Quoted 4 lines
I am using your tree origin/pm-wip/omap_device branch. On this I see a commit (
e89a95db7095d998991c564a756a33ee0ec722c5) in which a warning is printed inomap2_init_clk_clkdmif there is no clockdomain associated with a clock node.This is to check for the case of a missing clkdm when there should be one. You're trying to solve the problem of when there is no clockdomain for a given clock.
Quoted 2 lines
My idea was this warning plus the check so as to avoid the system crash would suffice.
I still think this should be flagged explicitly in the clock/hwmod, then check using the flag instead of of using NULL pointer check.
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
For readability, it makes it more clear that this is checking for some exceptions and not simply doing error checking.
Agreed. We could add a flag and base the check on the flag. But still if some clock node does not have a clkdomain and forgets to put this flag, the system will still crash.
Yes, but with a clear message as to why.
Quoted 1 line
So have have both the checks ??
I don't think so, because the check you added will allow the system to continue in a known non-working state.
Kevin
Quoted 56 lines
Quoted 56 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 10:36 PM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.Quoted 6 lines · Thara
Quoted 6 lines
Quoted 6 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 4:39 AM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change thesleep
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
dependencies.
Quoted 12 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Are you suggesting adding a flag to the clock node in case it does not have an associated clockdomain?
Yes, or to the hwmod.
Quoted 4 lines
I am using your tree origin/pm-wip/omap_device branch. On this I see a commit (
e89a95db7095d998991c564a756a33ee0ec722c5) in which a warning is printed inomap2_init_clk_clkdmif there is no clockdomain associated with a clock node.This is to check for the case of a missing clkdm when there should be one. You're trying to solve the problem of when there is no clockdomain for a given clock.
Quoted 5 lines
Quoted 5 lines
Quoted 2 lines
My idea was this warning plus the check so as to avoid the system crash would suffice.
I still think this should be flagged explicitly in the clock/hwmod, then check using the flag instead of of using NULL pointer check.
Two points. First , if there is a flag it should be in clock node. Hwmod structure should not bother with whether the clock node has a clockdomain or not. Second null pointer checks are there every where in the kernel and generally is a good programming practice.
Quoted 5 lines
Quoted 5 lines
For readability, it makes it more clear that this is checking for some exceptions and not simply doing error checking.
Kevin
Quoted 54 lines · Thara
Quoted 54 lines
Quoted 54 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 10:36 PM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.Quoted 6 lines · Thara
Quoted 6 lines
Quoted 6 lines
-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, December 10, 2009 4:39 AM
To: Gopinath, Thara
Cc: [email protected]; Paul Walmsley
Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change thesleep
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
dependencies.
Quoted 12 lines · Thara Gopinath
Some clock nodes like
wdt1_fck,sr1_fck,sr2_fcketc do not have a clock domain associated . For such nodes accessing the clock domain pointers in_add_initiator_depand_del_initiator_dep, will lead to null pointer defreferencing crash. Adding support in these functions to check for existence of clkdm pointer before trying to acess it. Even if tomorrow we correct all the clock nodes to have an associated clock domain, checking for the existence of the pointer is a good programming practice.Signed-off-by Thara Gopinath <[email protected]>Cc Paul Walmsley <[email protected]>---
This patch depends on http://patchwork.kernel.org/patch/63383/Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
I think clocks without associated clockdomains should be handled with a flag instead of a check for a NULL pointer.
Your current approach will silently fail if someone forgets to add a clockdomain to a clock that should have one.
Are you suggesting adding a flag to the clock node in case it does not have an associated clockdomain?
Yes, or to the hwmod.
Quoted 4 lines
I am using your tree origin/pm-wip/omap_device branch. On this I see a commit (
e89a95db7095d998991c564a756a33ee0ec722c5) in which a warning is printed inomap2_init_clk_clkdmif there is no clockdomain associated with a clock node.This is to check for the case of a missing clkdm when there should be one. You're trying to solve the problem of when there is no clockdomain for a given clock.
Quoted 7 lines
Quoted 5 lines
Quoted 5 lines
Quoted 2 lines
My idea was this warning plus the check so as to avoid the system crash would suffice.
I still think this should be flagged explicitly in the clock/hwmod, then check using the flag instead of of using NULL pointer check.
Two points. First , if there is a flag it should be in clock node.
OK
Quoted 4 lines
Hwmod structure should not bother with whether the clock node has a clockdomain or not.
Second null pointer checks are there every where in the kernel and generally is a good programming practice.
Of course, NULL pointer checks are commonly used to catch errors. My point is is that there are two things we're trying to catch here, and I was hoping to distinguish between the two:
1) errors: clocks that *have* a clockdomain but it's missing from the struct clk:
2) exceptions: clocks that do not have an associated clockdomain and need special treatment
IMHO, for readability, these should not be handled with the same NULL pointer check. Using NULL pointer check to catch (1) is perfectly OK (and what I did with my WARN patch.)
My point (and my opinion) is only that (2) should not be handled by the same NULL pointer check, but rather by using a flag to indicate the special case.
Kevin