[vbox-dev] Removing code only used by no longer supported hosts (was Opportunity to significantly shrink the vboxsf Linux driver)

[Hans de Goede]

Hi,

On 10-08-17 16:11, Hans de Goede wrote:
> Hi,
> 
> On 08-08-17 16:23, Hans de Goede wrote:
>> Hi,
>>
>> On 07/25/2017 11:06 AM, Michael Thayer wrote:
>>  > 13.07.2017 16:42, Michael Thayer wrote:
>>  >> 10.07.2017 12:29, Hans de Goede wrote:
>>  >> [Patch to shared folder code proposed by Hans to reduce the line count
>>  >> of the Linux driver.]
>>  >>
>>  >> I reviewed the patch and made a few changes.  Not yet tested on any
>>  >> platform, but I am posting my adjusted version so that other people can
>>  >> test, including on FreeBSD and Haiku which I cannot do easily.  Note
>>  >> that this applies to the normal VirtualBox tree, not to the packaged
>>  >> Linux driver.
>>  > I redid the patch again, after realising that the non-physical-page path
>>  > was for supporting VirtualBox 3.0 and older on the host, which we have
>>  > not tested or supported for a long time.  Therefore
>>  > VbglR0CanUsePhysPageList() could be removed altogether.
>> Ok, so I assume that similar checks to use physical page-lists can
>> be removed from vboxguest too ?
>>
>> And talking about removing support for old features, are hosts
>> without intel-vt or the amd equivalent still supported; if not then
>> I think that vgdrvInitFixateGuestMappings can be removed ?
>>
>> And what about vgdrvReportGuestInfo() can that be modified to
>> simply always first call VMMDevReq_ReportGuestInfo2 and then
>> VMMDevReq_ReportGuestInfo or are some of the paths checking for
>> rc == VERR_NOT_SUPPORTED || rc == VERR_NOT_IMPLEMENTED still
>> needed ?
> 
> First of all thank you for the answers to the above questions (see archives).
> 
> I've just found one more piece of code where I wonder if that should
> be kept. VbglR0SfMapFolder first tries SHFL_FN_MAP_FOLDER and if thar
> fails falls back to SHFL_FN_MAP_FOLDER_OLD. Are hosts only supporting
> SHFL_FN_MAP_FOLDER_OLD still supported, or can the fallback path
> be dropped ?

Ping? Getting an answer to this would be goold.

Likewise vfsmod.c has this:

         if (info->nullchar != '\0'
             || info->signature[0] != VBSF_MOUNT_SIGNATURE_BYTE_0
             || info->signature[1] != VBSF_MOUNT_SIGNATURE_BYTE_1
             || info->signature[2] != VBSF_MOUNT_SIGNATURE_BYTE_2) {
                 /*
                  * An old version of mount.vboxsf made the syscall.
                  * Translate the old parameters to the new structure.
                  */
                 struct vbsf_mount_info_old *info_old = (void *)info;
                 static struct vbsf_mount_info_new info_compat;

Is this support for older mount.vboxsf binaries still necessary ?

Regards,

Hans