Skip to contentSkip to conversation

RFC: introduce malloc_cleared() ?

U-Boot Reinhard Meyer 4 participants 8 messages Oct 14, 2010

lore
Share & tools
lore
#0 lore raw

Hello,

instead of adding or having the sequence of

malloc(), memset()

in all places where a definitely sane initialized structure is required it would make more sense to introduce a single function doing both. (The standard "calloc()" does not suit well here.) Overall that should even slightly decrease code size.

Reinhard

#1 lore raw

Dear Reinhard Meyer,

Quoted 9 lines · In message <[email protected]> you wrote:

instead of adding or having the sequence of

malloc(), memset()

in all places where a definitely sane initialized structure is required it would make more sense to introduce a single function doing both. (The standard "calloc()" does not suit well here.) Overall that should even slightly decrease code size.

And why exactly does calloc() not fit?

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
A verbal contract isn't worth the paper it's written on.
                                                    -- Samuel Goldwyn
#2 lore raw
Quoted 5 lines · Wolfgang Denk

Dear Reinhard Meyer,

Quoted 2 lines · In message <[email protected]> you wrote:

instead of adding or having the sequence of

Quoted 8 lines
Quoted 6 lines

malloc(), memset()

in all places where a definitely sane initialized structure is required it would make more sense to introduce a single function doing both. (The standard "calloc()" does not suit well here.) Overall that should even slightly decrease code size.

And why exactly does calloc() not fit?

It has two parameters... Produces more code to always supply an extra "1"...

I am game with using calloc(). That's all.

But for malloc() in such driver init situations, another code saving function with error message would be more effective. I don't see any possible way of continuing u-boot when malloc() for a relatively small structure already fails, and a (apparently) needed driver cannot be initialized.

I am not talking about malloc()'s for buffers or similar where continuing of u-boot might be possible.

Reinhard

#3 lore raw
Quoted 21 lines
Quoted 5 lines · Wolfgang Denk

Dear Reinhard Meyer,

Quoted 2 lines · In message <[email protected]> you wrote:

instead of adding or having the sequence of

Quoted 8 lines
Quoted 6 lines

malloc(), memset()

in all places where a definitely sane initialized structure is required it would make more sense to introduce a single function doing both. (The standard "calloc()" does not suit well here.) Overall that should even slightly decrease code size.

And why exactly does calloc() not fit?

It has two parameters... Produces more code to always supply an extra "1"...

I am game with using calloc(). That's all.

Quoted 7 lines

But for malloc() in such driver init situations, another code saving function with error message would be more effective. I don't see any possible way of continuing u-boot when malloc() for a relatively small structure already fails, and a (apparently) needed driver cannot be initialized.

I am not talking about malloc()'s for buffers or similar where

continuing

Quoted 1 line

of u-boot might be possible.

I would borrow the naming from linux and call it mallocz

Jocke

#4 lore raw

Dear Joakim Tjernlund,

Quoted 2 lines · In message <OF32166834.809C7D71-ONC12577BC.0031112B-C12577BC.00313F63@transmode…

I would borrow the naming from linux and call it mallocz

... except that Linux does not use any "mallocz" anywhere :-)

I guess you mean [k]zalloc() ?

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
When a program is being  tested,  it  is  too  late  to  make  design
changes.                  -- Geoffrey James, "The Tao of Programming"
#5 lore raw

Wolfgang Denk <[email protected]> wrote on 2010/10/14 11:38:57:

Quoted 11 lines

Dear Joakim Tjernlund,

In message <OF32166834.809C7D71-ONC12577BC.0031112B-C12577BC.

Quoted 2 lines · 00313F63 at transmode.se> you wrote:

I would borrow the naming from linux and call it mallocz

... except that Linux does not use any "mallocz" anywhere :-)

I guess you mean [k]zalloc() ?

Yes, memory played tricks on me :)

#6 lore raw
In reply to Reinhard Meyer
Quoted 18 lines · Reinhard Meyer
Quoted 5 lines · Wolfgang Denk

Dear Reinhard Meyer,

Quoted 2 lines · In message <[email protected]> you wrote:

instead of adding or having the sequence of

Quoted 8 lines
Quoted 6 lines

malloc(), memset()

in all places where a definitely sane initialized structure is required it would make more sense to introduce a single function doing both. (The standard "calloc()" does not suit well here.) Overall that should even slightly decrease code size.

And why exactly does calloc() not fit?

It has two parameters... Produces more code to always supply an extra "1"...

Quoted 7 lines

I am game with using calloc(). That's all.

But for malloc() in such driver init situations, another code saving function with error message would be more effective. I don't see any possible way of continuing u-boot when malloc() for a relatively small structure already fails, and a (apparently) needed driver cannot be initialized.

The failed malloc can cause the dailed driver to unload malloc'd memory and print a warning to the console. User intervention (changing environment variables to avoid the out-of-memory on reboot for example) can still be possible even though the driver failed.

Quoted 4 lines

I am not talking about malloc()'s for buffers or similar where continuing of u-boot might be possible.

Regards,

Graeme

#7 lore raw
In reply to Reinhard Meyer

Dear Reinhard Meyer,

Quoted 5 lines · In message <[email protected]> you wrote:
Quoted 2 lines

And why exactly does calloc() not fit?

It has two parameters... Produces more code to always supply an extra "1"...

Ah.

Quoted 2 lines

But for malloc() in such driver init situations, another code saving function with error message would be more effective. I don't see any

Would it? You still need to check the return code of the allocation function then, because you want to breeak out of the caller.

Quoted 3 lines

possible way of continuing u-boot when malloc() for a relatively small structure already fails, and a (apparently) needed driver cannot be initialized.

Maybe, maybe not. As mentioned before: network access may not be so important at all.

Quoted 2 lines

I am not talking about malloc()'s for buffers or similar where continuing of u-boot might be possible.

Continuing U-Boot without network access is also possible. If someone pulls the cable you also continue, or don't you?

Best regards,

Wolfgang Denk

Signature · 7 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"Pardon me for breathing, which I never do anyway so I don't know why
I bother to say it, oh God, I'm so depressed. Here's another of those
self-satisfied doors. Life! Don't talk to me about life."
                                        - Marvin the Paranoid Android