On 18.01.2012 15:36, Phillip Susi wrote:
On 1/18/2012 8:23 AM, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
I disagree with this patch being good. msdos layout has no inherent
limit on number of logical partitions. The above mentioned limit comes
from how Linux allocates minor SCSI bits and presence of this limit is
Linux bug. non-SCSI disks on Linux have higher limit. Even on SCSI disks
it can be circumvented with kpartx. Moreover other kernels and tools are
probably not affected by this limit.
If you really want to handle this limit AFAIK it's not specific to just
msdos partitions, so it has to be on higher level and happen only on
SCSI disks and only when running under Linux or similarly affected
kernels (e.g. kOpenSolaris has a similar limit) and has to be reduced to
a warning since a bug in some other software isn't a reason to create a
same bug.
While it is true that there is no limit imposed by the partition table,
for the purposes of parted there must be a limit somewhere so it can
decide how many partitions it needs to try and remove from the kernel
partition table.
Ideally there should be a function allowing to know how many entries
this table has.
If kernel API is unmodifiable and has inherent limit IMHO the best
strategy is to make this limit pertain only to kernel-specific parts.
The kernel allows up to 256 partitions so the limit
could be set that high, but since the limit was 16 before and nobody
complained, I figured that raising it to 64 should be plenty. I think
that anyone actually hitting that limit needs to have their head examined.
I'll see when I can get an appointment with the doctor. I've had
problems with hitting those limits when I created ultra-multi-boot disk
--
Regards
Vladimir 'φ-coder/phcoder' Serbinenko
