Hi Brian, Thanks for your comments, we need internal discussion for your raised issues, will reply later this week.
Thanks, Eric > -----Original Message----- > From: Brian J. Johnson [mailto:bjohn...@sgi.com] > Sent: Friday, February 19, 2016 5:18 AM > To: Andrew Fish; Dong, Eric > Cc: El-Haj-Mahmoud, Samer; Bi, Dandan; edk2-devel@lists.01.org; Gao, Liming; > Rothman, Michael A; Hale, Robert P; Kinney, Michael D > Subject: Re: [edk2] [patch] MdeModulePkg: Make HII configuration settings > available to OS runtime > > On 02/17/2016 10:25 PM, Andrew Fish wrote: > > > >> On Feb 17, 2016, at 6:50 PM, Dong, Eric <eric.d...@intel.com> wrote: > >> > >> Hi All, > >> > >> Thanks for your comments. I add my explanation below: > >> > >> Only hook ReadyToBoot event is not enough. Different drivers may hook this > >> event and some may update string package in their > callback function. The order to call these callback functions is random, so > only hook this event may miss some changes. > >> > > Eric, > > Could you enable your HII exporting callbacks in a ReadyToBoot event? > Then you wouldn't needlessly pay the export cost each time a HII element > is added or changed as drivers are loaded. > > >> This feature is described in UEFI spec 2.6 chapter 31.2.11.1. It required > >> to export all HiiDatabae data and current configuration data. > > > > I don't see any language in that section that makes it required. It seems > > to just describe how to do it. > > > > Agreed. And section 31.2.1.10 says: > > ===== > When it is desired that the forms data be used in the presence of an > O/S, this specification describes > a means by which to support this capability. By being able to > encapsulate the data and export it > through standard means such that an O/S agent (e.g. forms > browser/processor) can retrieve it, O/S- > present usage models can be made available for further value-add > implementations. > ===== > > The wording implies that exporting is not always desired, and that > O/S-present usage of forms data is a value-add feature, which a platform > may choose to support or not. > > >> We add this feature because we strongly encourage the export of config > >> data to support manageability of platform config both for pre- > OS and OS runtime operations. > >> > > > > Generally the UEFI spec and the edk2 don't try to force platform policy. > > For example exposing some configuration information could be > considered a security vulnerability on some platforms, so it should not be > forced on a platform. > > > > Thanks, > > > > Andrew Fish > > > >> Also we collect the performance data from our reference platform(Detail > >> see below). The boot performance is small and we can > ignore it. This feature cost extra 869KB, but we think the size cost is not > care because current we use 16M size of bios flash. > >> > >> HiiDatabase driver data size cost(This feature add in HiiDatabase driver): > >> Without this feature: > >> Used data size: 151KB > >> With this feature > >> Used data size: 1020 KB > >> PS: This extra size cost mainly depend on how much Hii packages in your > >> platform. > >> > >> Boot Time cost: > >> Without this feature: > >> Boot Time cost: 1st:3.353s 2nd: 3.035s > >> With this feature: > >> Boot Time cost: 1st: 3.125s 2nd: 3.126s > >> > >> Base on the above analysis, we provide this design and prefer to always > >> enable this feature instead of use a PCD to control it. > >> > >> Thanks, > >> Eric > >> > > My particular concern involves a simulator (far slower than hardware) > for an extremely large, complex server platform with a large amount of > HII data. On this platform, we end up disabling the largest but least > used setup screens when running on the simulator, simply to avoid the > time overhead of processing the HII data. This saves many minutes of > execution time. > > >>> -----Original Message----- > >>> From: El-Haj-Mahmoud, Samer [mailto:samer.el-haj-mahm...@hpe.com] > >>> Sent: Wednesday, February 17, 2016 2:14 AM > >>> To: Andrew Fish; Brian J. Johnson > >>> Cc: Bi, Dandan; edk2-devel@lists.01.org; Dong, Eric; Gao, Liming > >>> Subject: RE: [edk2] [patch] MdeModulePkg: Make HII configuration settings > >>> available to OS runtime > >>> > >>> +1 > >>> > >>> I also would add there may be some HII strings that are hidden from user > >>> interfaces, and reflect settings for field service or > >>> troubleshooting, and that a mass export to the OS may expose these > >>> settings to OS runtime code and possibly applications. > >>> > >>> > >>> > >>> -----Original Message----- > >>> From: edk2-devel [mailto:edk2-devel-boun...@lists.01.org] On Behalf Of > >>> Andrew Fish > >>> Sent: Tuesday, February 16, 2016 10:37 AM > >>> To: Brian J. Johnson <bjohn...@sgi.com> > >>> Cc: Dandan Bi <dandan...@intel.com>; edk2-devel@lists.01.org; Eric Dong > >>> <eric.d...@intel.com>; Liming Gao > <liming....@intel.com> > >>> Subject: Re: [edk2] [patch] MdeModulePkg: Make HII configuration settings > >>> available to OS runtime > >>> > >>> > >>>> On Feb 16, 2016, at 8:33 AM, Brian J. Johnson <bjohn...@sgi.com> wrote: > >>>> > >>>> On 02/16/2016 12:58 AM, Dandan Bi wrote: > >>>>> This feature is aimed to allow OS make use of the HII database during > >>>>> runtime. In this case, the contents of the HII Database is exported > >>>>> to a buffer. The pointer to the buffer is placed in the EFI System > >>>>> Configuration Table, where it can be retrieved by an OS application. > >>>>> > >>>>> Export the configuration data and contents of HiiDatabase when driver > >>>>> add package, update package and remove package. > >>>>> For string and image may also need to update the contents of > >>>>> HiiDatabase when NewString/SetString, NewImage/SetImage. > >>>>> > >>>>> Cc: Liming Gao <liming....@intel.com> > >>>>> Cc: Eric Dong <eric.d...@intel.com> > >>>>> Contributed-under: TianoCore Contribution Agreement 1.0 > >>>>> Signed-off-by: Dandan Bi <dandan...@intel.com> ... > >>>> > >>>> Please make this behavior selectable via a PCD. HII operations are very > >>>> expensive, especially on simulators. I don't want to pay the > >>> export time every time a package is added or string is changed. Also, > >>> platforms should be able to decide if they want to offer this > data to > >>> the OS. > >>>> > >>> > >>> +1 > >>> > >>> I would want to opt out to NOT take the memory away from the OS if the > >>> platform did not care about the feature. > >>> > >>> Thanks, > >>> > >>> Andrew Fish > >>> > >>>> Why not just export the data once, using a "ready to boot" event hook? > >>>> > >>>> Thanks, > >>>> -- > >>>> > >>>> Brian J. Johnson > >>>> > >>>> -------------------------------------------------------------------- > >>>> > >>>> My statements are my own, are not authorized by SGI, and do not > >>>> necessarily represent SGI's positions. > >>>> _______________________________________________ > >>>> edk2-devel mailing list > >>>> edk2-devel@lists.01.org > >>>> https://lists.01.org/mailman/listinfo/edk2-devel > >>> > >>> _______________________________________________ > >>> edk2-devel mailing list > >>> edk2-devel@lists.01.org > >>> https://lists.01.org/mailman/listinfo/edk2-devel > >> > > > > > -- > > Brian > > -------------------------------------------------------------------- > > "If you want your dreams to come true: Wake Up!" > -- Jaime Roman's Radio _______________________________________________ edk2-devel mailing list edk2-devel@lists.01.org https://lists.01.org/mailman/listinfo/edk2-devel
