Host protected area on suspend/resume

Host protected area on suspend/resume

[Matthew Garrett]

On boot, Linux will attempt to disable the host protected area on a 
disk. After a suspend/resume cycle, the BIOS may reenable it (seen on a 
Thinkpad T40 and R40). As a result, the kernel is now unable to access 
the HPA.

Is there any issue with just adding a call to idedisk_check_hpa() in the 
IDE resume code?
-- 
Matthew Garrett | mjg59@srcf.ucam.org

Re: Host protected area on suspend/resume

[Tejun Heo]

Matthew Garrett wrote:
> On boot, Linux will attempt to disable the host protected area on a 
> disk. After a suspend/resume cycle, the BIOS may reenable it (seen on a 
> Thinkpad T40 and R40). As a result, the kernel is now unable to access 
> the HPA.
> 
> Is there any issue with just adding a call to idedisk_check_hpa() in the 
> IDE resume code?

  This has come up several times now.  One thing I'm curious about is 
why we are disabling HPA on boot without consent from the user.  AFAIK, 
HPA is mostly used to implement hidden recovery/suspend storage areas 
and disabling automatically on boot increases the likeliness of 
destroying them.  What do we gain by disabling HPA on boot?  Are there 
some dumb machines which unnecessarily sets HPA and reduces the capacity 
of drives excessively?  Even in such cases, wouldn't it be better to do 
idedisk_check_hpa() only when kernel parameter explicitly says so?

-- 
tejun

Re: Host protected area on suspend/resume

[Matthew Garrett]

On Tue, Jul 12, 2005 at 11:49:35PM +0900, Tejun Heo wrote:

>  This has come up several times now.  One thing I'm curious about is 
> why we are disabling HPA on boot without consent from the user.  AFAIK, 
> HPA is mostly used to implement hidden recovery/suspend storage areas 
> and disabling automatically on boot increases the likeliness of 
> destroying them.  What do we gain by disabling HPA on boot?  Are there 
> some dumb machines which unnecessarily sets HPA and reduces the capacity 
> of drives excessively?  Even in such cases, wouldn't it be better to do 
> idedisk_check_hpa() only when kernel parameter explicitly says so?

I'm not sure why we're doing it, but reverting this behaviour is likely 
to make some systems unbootable (install with kernel which disables HPA, 
format entire drive, put filesystem on it).
-- 
Matthew Garrett | mjg59@srcf.ucam.org

Re: Host protected area on suspend/resume

[Tejun Heo]

Matthew Garrett wrote:
> On Tue, Jul 12, 2005 at 11:49:35PM +0900, Tejun Heo wrote:
> 
> 
>> This has come up several times now.  One thing I'm curious about is 
>>why we are disabling HPA on boot without consent from the user.  AFAIK, 
>>HPA is mostly used to implement hidden recovery/suspend storage areas 
>>and disabling automatically on boot increases the likeliness of 
>>destroying them.  What do we gain by disabling HPA on boot?  Are there 
>>some dumb machines which unnecessarily sets HPA and reduces the capacity 
>>of drives excessively?  Even in such cases, wouldn't it be better to do 
>>idedisk_check_hpa() only when kernel parameter explicitly says so?
> 
> 
> I'm not sure why we're doing it, but reverting this behaviour is likely 
> to make some systems unbootable (install with kernel which disables HPA, 
> format entire drive, put filesystem on it).

  Yeah... it will immediately make disks unusable.  sad, sad.

-- 
tejun

Re: Host protected area on suspend/resume

[Greg Freemyer]

On 7/12/05, Tejun Heo wrote:
> Matthew Garrett wrote:
> > On boot, Linux will attempt to disable the host protected area on a
> > disk. After a suspend/resume cycle, the BIOS may reenable it (seen on a
> > Thinkpad T40 and R40). As a result, the kernel is now unable to access
> > the HPA.
> >
> > Is there any issue with just adding a call to idedisk_check_hpa() in the
> > IDE resume code?
> 
>   This has come up several times now.  One thing I'm curious about is
> why we are disabling HPA on boot without consent from the user.  AFAIK,
> HPA is mostly used to implement hidden recovery/suspend storage areas
> and disabling automatically on boot increases the likeliness of
> destroying them.  What do we gain by disabling HPA on boot?  Are there
> some dumb machines which unnecessarily sets HPA and reduces the capacity
> of drives excessively?  Even in such cases, wouldn't it be better to do
> idedisk_check_hpa() only when kernel parameter explicitly says so?
> 
> --
> tejun

Crawling out from under my rock, I agree with Tejun.

Always enabling access to HPA seems very strange.

It also eliminates the ability to use the IDE Offset feature.  IIUC,
the offset feature is a boolean flag that, when true, basically tells
the drive that all LBA references are into the HPA area.

By using the HPA in conjunction with the Offset flag, one has the
ability to have two complete disk layouts.  Each could have its own
boot sector, partition table, and partitions.  I have never used this
feature, but I do have DOS based tools that supposidely control the
HPA config. and the state of the offset boolean.

I routinely perform "forensic image aquires".  Normally that is a
simple "dd if=/dev/hdx.....".

For me the untlimate would be to have Linux support for something like:
disable offset
dd normal sectors
enable offset
dd hpa sectors

That way I capture the areas as seperate images and if 2 boot sectors,
partition tables, partitions etc. exist I can analyse them seperately.

Conclusion, automatically overriding HPA seems to be outside of the
normal Linux way of doing things.  At a minimum, there should be a
flag to prevent it from happening.  If it is needed there are existing
userspace linux tools that can modify the HPA config.


FYI: In addition to HPA potentially reducing usable disk size, the
size can also be artificially reduced via DCO.  I am not aware of any
Linux tools for addressing/eliminating artificial DCO restrictions.

Greg
-- 
Greg Freemyer
The Norcross Group
Forensics for the 21st Century

Re: Host protected area on suspend/resume

[Grant Coady]

On Tue, 12 Jul 2005 23:49:35 +0900, Tejun Heo <htejun@gmail.com> wrote:

>Matthew Garrett wrote:
>> On boot, Linux will attempt to disable the host protected area on a 
>> disk. After a suspend/resume cycle, the BIOS may reenable it (seen on a 
...
>  This has come up several times now.  One thing I'm curious about is 
>why we are disabling HPA on boot without consent from the user.  AFAIK, 
>HPA is mostly used to implement hidden recovery/suspend storage areas 
>and disabling automatically on boot increases the likeliness of 
>destroying them.  What do we gain by disabling HPA on boot?

It simply implies one _must_not_ use linux-kernel 2.6 to prepare IDE 
drives, stay with 2.4 series for drive preparation until the bug is 
fixed.

I'd noticed the issue but ignored it as I still setup systems with 
2.4 prior to installing 2.6.

--Grant.

Re: Host protected area on suspend/resume

[Sergey Vlasov]

[-- Attachment #1: Type: text/plain, Size: 1279 bytes --]

On Tue, 12 Jul 2005 23:49:35 +0900 Tejun Heo wrote:

> Matthew Garrett wrote:
> > On boot, Linux will attempt to disable the host protected area on a 
> > disk. After a suspend/resume cycle, the BIOS may reenable it (seen on a 
> > Thinkpad T40 and R40). As a result, the kernel is now unable to access 
> > the HPA.
> > 
> > Is there any issue with just adding a call to idedisk_check_hpa() in the 
> > IDE resume code?
> 
>   This has come up several times now.  One thing I'm curious about is 
> why we are disabling HPA on boot without consent from the user.  AFAIK, 
> HPA is mostly used to implement hidden recovery/suspend storage areas 
> and disabling automatically on boot increases the likeliness of 
> destroying them.  What do we gain by disabling HPA on boot?  Are there 
> some dumb machines which unnecessarily sets HPA and reduces the capacity 
> of drives excessively?  Even in such cases, wouldn't it be better to do 
> idedisk_check_hpa() only when kernel parameter explicitly says so?

Support for disabling HPA during boot is needed mainly for old
machines with BIOS limitations. Some disks have jumpers which limit
capacity to 32 GB, so that such old BIOSes can recognize the disk
without hanging at boot; disabling HPA gives access to the whole
capacity.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]