[PATCH 1/2] Remove kernel version check

[Adhemerval Zanella]

On 28/02/2022 16:00, Florian Weimer wrote:
> * Adhemerval Zanella via Libc-alpha:
> 
>> The kernel version check is used to avoid glibc to run on older
>> kernels where some syscall are not available and fallback code are
>> not enabled to handle graciously fail.  However, it does not prevent
>> if the kernel does not correctly advertise its version through
>> vDSO note, uname or procfs.
>>
>> Also kernel version checks are sometime not desirable by users,
>> where they want to deploy on different system with different kernel
>> version knowing the minimum set of syscall is always presented on
>> such systems.
>>
>> The kernel version check has been removed along with the
>> LD_ASSUME_KERNEL environment variable.  The minimum kernel used to
>> built glibc is still provided through NT_GNU_ABI_TAG ELF note and
>> also printed when libc.so is issued.
> 
> Missing Linux: prefix in the commit subject, I would.
> 
>> diff --git a/NEWS b/NEWS
>> index 626eeabf5d..bbbcc55b0c 100644
>> --- a/NEWS
>> +++ b/NEWS
>> @@ -17,6 +17,11 @@ Deprecated and removed features, and other changes affecting compatibility:
>>    removal of the LD_TRACE_PRELINKING, and LD_USE_LOAD_BIAS, environment
>>    variables and their functionality in the dynamic loader.
>>  
>> +* The kernel version check has been removed along with the LD_ASSUME_KERNEL
>> +  environment variable.  The minimum kernel used to built glibc is still
>> +  provided through NT_GNU_ABI_TAG ELF note and also printed when libc.so
>> +  is issued directly.
> 
> Mention “Linux” here?

I changed to "The Linux kernel version check ..."

> 
>> index 892e8ef2f6..ba175a0035 100644
>> --- a/elf/dl-load.c
>> +++ b/elf/dl-load.c
>> @@ -1631,7 +1631,6 @@ open_verify (const char *name, int fd,
>>        ElfW(Phdr) *phdr, *ph;
>>        ElfW(Word) *abi_note;
>>        ElfW(Word) *abi_note_malloced = NULL;
>> -      unsigned int osversion;
>>        size_t maplength;
>>  
>>        /* We successfully opened the file.  Now verify it is a file
>> @@ -1695,13 +1694,16 @@ open_verify (const char *name, int fd,
>>  #endif
>>  	      )
>>  	    errstring = N_("invalid ELF header");
>> +
>>  	  else if (ehdr->e_ident[EI_CLASS] != ELFW(CLASS))
>>  	    {
>>  	      /* This is not a fatal error.  On architectures where
>>  		 32-bit and 64-bit binaries can be run this might
>>  		 happen.  */
>>  	      *found_other_class = true;
>> -	      goto close_and_out;
>> +	      __close_nocancel (fd);
>> +	      __set_errno (ENOENT);
>> +	      return -1;
>>  	    }
>>  	  else if (ehdr->e_ident[EI_DATA] != byteorder)
>>  	    {
>> @@ -1736,7 +1738,11 @@ open_verify (const char *name, int fd,
>>  	  goto lose;
>>  	}
>>        if (! __glibc_likely (elf_machine_matches_host (ehdr)))
>> -	goto close_and_out;
>> +	{
>> +	  __close_nocancel (fd);
>> +	  __set_errno (ENOENT);
>> +	  return -1;
>> +	}
>>        else if (__glibc_unlikely (ehdr->e_type != ET_DYN
>>  				 && ehdr->e_type != ET_EXEC))
>>  	{
>> @@ -1768,7 +1774,11 @@ open_verify (const char *name, int fd,
>>        if (__glibc_unlikely (elf_machine_reject_phdr_p
>>  			    (phdr, ehdr->e_phnum, fbp->buf, fbp->len,
>>  			     loader, fd)))
>> -	goto close_and_out;
>> +	{
>> +	  __close_nocancel (fd);
>> +	  __set_errno (ENOENT);
>> +	  return -1;
>> +	}
>>  
>>        /* Check .note.ABI-tag if present.  */
>>        for (ph = phdr; ph < &phdr[ehdr->e_phnum]; ++ph)
>> @@ -1820,18 +1830,6 @@ open_verify (const char *name, int fd,
>>  	    if (size == 0)
>>  	      continue;
>>  
>> -	    osversion = (abi_note[5] & 0xff) * 65536
>> -			+ (abi_note[6] & 0xff) * 256
>> -			+ (abi_note[7] & 0xff);
>> -	    if (abi_note[4] != __ABI_TAG_OS
>> -		|| (GLRO(dl_osversion) && GLRO(dl_osversion) < osversion))
>> -	      {
>> -	      close_and_out:
>> -		__close_nocancel (fd);
>> -		__set_errno (ENOENT);
>> -		fd = -1;
>> -	      }
>> -
>>  	    break;
>>  	  }
>>        free (abi_note_malloced);
> 
> Hmm.  I think you can delete the entire program header loop now?
> Nothing reads abi_note after these changes.

Indeed, I will removed it as well.

> 
> And there's now an unused
> 
> unsigned int _dl_osversion;
> 
> in elf/dl-support.c.

Ack.

> 
> I assume you want to handle the ldconfig changes separately?  I think we
> discussed that during Monday's call last week.

Yes, I forget to clean up this up. I will add another patch to do so.