Messages in this thread | | | Subject | Re: [PATCH 1/2] kernel/bpf: change "char *" string form to "char []" | From | liqiong <> | Date | Fri, 13 May 2022 10:04:30 +0800 |
| |
在 2022年05月13日 04:59, Daniel Borkmann 写道: > On 5/12/22 7:08 PM, liqiong wrote: >> 在 2022年05月12日 23:16, Yonghong Song 写道: >>> >>> On 5/12/22 7:28 AM, liqiong wrote: >>>> The string form of "char []" declares a single variable. It is better >>>> than "char *" which creates two variables. >>> >>> Could you explain in details about why it is better in generated codes? >>> It is not clear to me why your patch is better than the original code. >> >> The string form of "char *" creates two variables in the final assembly output, >> a static string, and a char pointer to the static string. Use "objdump -S -D *.o", >> can find out the static string occurring at "Contents of section .rodata". > > There are ~360 instances of this type in the tree from a quick grep, do you > plan to convert all them ?
Hi daniel,
I have fixed all the string form in kernel tree, summited two patches. Have searched the kernel tree by "grep -nHre char.*\*.*=.*\"", and checked all the "char *foo = "bar" " string form, only five instances are needed to fix in bpf direcotry (2 files) and trace directory (2 files). In most cases, need a char pointer anyway, just like this:
[const] char *foo = "bar"; if (xxx) foo = "blash";
In this situation, can't change "char *foo" to "char foo[]".
This work was published in "KernelJanitors/Todo". So, I fixed.
Thanks.
> > Thanks, > Daniel
| |