MdeModulePkg/UefiBootManagerLib: recursive call of BmRepairAllControllers

MdeModulePkg/UefiBootManagerLib: recursive call of BmRepairAllControllers

[Guo Heyi]

Hi folks,

In BmDriverHealth.c, function BmRepairAllControllers may recursively call itself
if some driver health protocol returns EfiDriverHealthStatusReconnectRequired.
However, if there is something wrong in some 3rd party driver (e.g. PCI oprom),
the driver health protocol of that driver may always return such status even
after one and another reconnect. The endless iteration will cause stack overflow
and then system exception, and it may be not easy to find that the exception is
actually caused by stack overflow.

So does it make sense to set a maximum count of this recursive call to avoid
whole system hang even there is a buggy 3rd party driver?

Thanks,

Gary (Heyi Guo)

Re: MdeModulePkg/UefiBootManagerLib: recursive call of BmRepairAllControllers

[Ni, Ruiyu]

On 2/24/2018 2:23 PM, Guo Heyi wrote:
> Hi folks,
> 
> In BmDriverHealth.c, function BmRepairAllControllers may recursively call itself
> if some driver health protocol returns EfiDriverHealthStatusReconnectRequired.
> However, if there is something wrong in some 3rd party driver (e.g. PCI oprom),
> the driver health protocol of that driver may always return such status even
> after one and another reconnect. The endless iteration will cause stack overflow
> and then system exception, and it may be not easy to find that the exception is
> actually caused by stack overflow.

yes. I agree.
If a buggy PCI oprom always return ReconnectRequired, the stack overflow
will happen.

How about we choose maximum recursive call depth as 10 and continue to
boot if call depth reaches 11?

> 
> So does it make sense to set a maximum count of this recursive call to avoid
> whole system hang even there is a buggy 3rd party driver?
> 
> Thanks,
> 
> Gary (Heyi Guo)
> 


-- 
Thanks,
Ray

Re: MdeModulePkg/UefiBootManagerLib: recursive call of BmRepairAllControllers

[Guo Heyi]

On Sat, Feb 24, 2018 at 04:20:52PM +0800, Ni, Ruiyu wrote:
> On 2/24/2018 2:23 PM, Guo Heyi wrote:
> >Hi folks,
> >
> >In BmDriverHealth.c, function BmRepairAllControllers may recursively call itself
> >if some driver health protocol returns EfiDriverHealthStatusReconnectRequired.
> >However, if there is something wrong in some 3rd party driver (e.g. PCI oprom),
> >the driver health protocol of that driver may always return such status even
> >after one and another reconnect. The endless iteration will cause stack overflow
> >and then system exception, and it may be not easy to find that the exception is
> >actually caused by stack overflow.
> 
> yes. I agree.
> If a buggy PCI oprom always return ReconnectRequired, the stack overflow
> will happen.
> 
> How about we choose maximum recursive call depth as 10 and continue to
> boot if call depth reaches 11?

That's fine for me :)

Thanks,

Gary (Heyi Guo)

> 
> >
> >So does it make sense to set a maximum count of this recursive call to avoid
> >whole system hang even there is a buggy 3rd party driver?
> >
> >Thanks,
> >
> >Gary (Heyi Guo)
> >
> 
> 
> -- 
> Thanks,
> Ray

Re: MdeModulePkg/UefiBootManagerLib: recursive call of BmRepairAllControllers

[Ni, Ruiyu]

Will you submit a patch for this change?

Thanks/Ray

> -----Original Message-----
> From: Guo Heyi [mailto:heyi.guo@linaro.org]
> Sent: Saturday, February 24, 2018 4:29 PM
> To: Ni, Ruiyu <ruiyu.ni@intel.com>
> Cc: Guo Heyi <heyi.guo@linaro.org>; edk2-devel@lists.01.org; Zeng, Star
> <star.zeng@intel.com>; Dong, Eric <eric.dong@intel.com>
> Subject: Re: MdeModulePkg/UefiBootManagerLib: recursive call of
> BmRepairAllControllers
> 
> On Sat, Feb 24, 2018 at 04:20:52PM +0800, Ni, Ruiyu wrote:
> > On 2/24/2018 2:23 PM, Guo Heyi wrote:
> > >Hi folks,
> > >
> > >In BmDriverHealth.c, function BmRepairAllControllers may recursively
> > >call itself if some driver health protocol returns
> EfiDriverHealthStatusReconnectRequired.
> > >However, if there is something wrong in some 3rd party driver (e.g.
> > >PCI oprom), the driver health protocol of that driver may always
> > >return such status even after one and another reconnect. The endless
> > >iteration will cause stack overflow and then system exception, and it
> > >may be not easy to find that the exception is actually caused by stack
> overflow.
> >
> > yes. I agree.
> > If a buggy PCI oprom always return ReconnectRequired, the stack
> > overflow will happen.
> >
> > How about we choose maximum recursive call depth as 10 and continue to
> > boot if call depth reaches 11?
> 
> That's fine for me :)
> 
> Thanks,
> 
> Gary (Heyi Guo)
> 
> >
> > >
> > >So does it make sense to set a maximum count of this recursive call
> > >to avoid whole system hang even there is a buggy 3rd party driver?
> > >
> > >Thanks,
> > >
> > >Gary (Heyi Guo)
> > >
> >
> >
> > --
> > Thanks,
> > Ray

Re: MdeModulePkg/UefiBootManagerLib: recursive call of BmRepairAllControllers

[Guo Heyi]

Sure.

On Sat, Feb 24, 2018 at 08:40:20AM +0000, Ni, Ruiyu wrote:
> Will you submit a patch for this change?
> 
> Thanks/Ray
> 
> > -----Original Message-----
> > From: Guo Heyi [mailto:heyi.guo@linaro.org]
> > Sent: Saturday, February 24, 2018 4:29 PM
> > To: Ni, Ruiyu <ruiyu.ni@intel.com>
> > Cc: Guo Heyi <heyi.guo@linaro.org>; edk2-devel@lists.01.org; Zeng, Star
> > <star.zeng@intel.com>; Dong, Eric <eric.dong@intel.com>
> > Subject: Re: MdeModulePkg/UefiBootManagerLib: recursive call of
> > BmRepairAllControllers
> > 
> > On Sat, Feb 24, 2018 at 04:20:52PM +0800, Ni, Ruiyu wrote:
> > > On 2/24/2018 2:23 PM, Guo Heyi wrote:
> > > >Hi folks,
> > > >
> > > >In BmDriverHealth.c, function BmRepairAllControllers may recursively
> > > >call itself if some driver health protocol returns
> > EfiDriverHealthStatusReconnectRequired.
> > > >However, if there is something wrong in some 3rd party driver (e.g.
> > > >PCI oprom), the driver health protocol of that driver may always
> > > >return such status even after one and another reconnect. The endless
> > > >iteration will cause stack overflow and then system exception, and it
> > > >may be not easy to find that the exception is actually caused by stack
> > overflow.
> > >
> > > yes. I agree.
> > > If a buggy PCI oprom always return ReconnectRequired, the stack
> > > overflow will happen.
> > >
> > > How about we choose maximum recursive call depth as 10 and continue to
> > > boot if call depth reaches 11?
> > 
> > That's fine for me :)
> > 
> > Thanks,
> > 
> > Gary (Heyi Guo)
> > 
> > >
> > > >
> > > >So does it make sense to set a maximum count of this recursive call
> > > >to avoid whole system hang even there is a buggy 3rd party driver?
> > > >
> > > >Thanks,
> > > >
> > > >Gary (Heyi Guo)
> > > >
> > >
> > >
> > > --
> > > Thanks,
> > > Ray