Skip to contentSkip to conversation

iBFT features for v2.6.36

GIT PULL Konrad Rzeszutek 8 participants 12 messages Aug 2, 2010 – Aug 4, 2010

lore
Share & tools
lore
#0 lore raw

Hi Linus,

This pull requests consists of features to the iBFT driver. Specifically:

- support v1.03 of the spec. The spec in question is available at:
ftp://ftp.software.ibm.com/systems/support/bladecenter/iscsi_boot_firmware_table_v1.03.pdf
- separate the iBFT parsing from the SysFS interface.

These patches have been through a few rounds of review and have been in linux-next for two months. James Bottomley is using this tree in his branch for the BNX2 driver, so please merge these patches before his.

It's all at:

git://git.kernel.org/pub/scm/linux/kernel/git/konrad/ibft-2.6.git for-linus

Thanks,

Konrad

Konrad Rzeszutek Wilk (1):
ibft: For UEFI machines actually do scan ACPI for iBFT.

Mike Christie (2):
ibft: separate ibft parsing from sysfs interface
ibft: convert iscsi_ibft module to iscsi boot lib

Peter Jones (2):
ibft: Update iBFT handling for v1.03 of the spec.
ibft: Use IBFT_SIGN instead of open-coding the search string.

 drivers/firmware/Kconfig            |    9 +
 drivers/firmware/Makefile           |    1 +
 drivers/firmware/iscsi_boot_sysfs.c |  481 +++++++++++++++++++++++
 drivers/firmware/iscsi_ibft.c       |  726 +++++++++++++----------------------
 drivers/firmware/iscsi_ibft_find.c  |   56 +++-
 include/linux/iscsi_boot_sysfs.h    |  123 ++++++
 include/linux/iscsi_ibft.h          |   12 +-
 7 files changed, 924 insertions(+), 484 deletions(-)

What is iBFT?
The iBFT is an equivalent to the Boot Flag, except that its geared towards iSCSI and hence requires much more information (such as the IP of target, passwords, which device to login, etc). iBFT is a data structure populated by the BIOS or the NIC to contain this so that the OS can read it and login to the iSCSI and present the boot device to the initrd for mounting / FS.

#1 lore raw
Quoted 7 lines · Konrad Rzeszutek

What is iBFT?
The iBFT is an equivalent to the Boot Flag, except that its geared towards iSCSI and hence requires much more information (such as the IP of target, passwords, which device to login, etc). iBFT is a data structure populated by the BIOS or the NIC to contain this so that the OS can read it and login to the iSCSI and present the boot device to the initrd for mounting / FS.

I really don't see iBFT as equivalent to the boot flag at all. The boot flag returns the status of the previous boot attempt; iBFT contains information about where to find the current root.

Unfortunately, we're increasingly seeing a proliferation of this kind of nonstandard ACPI tables, because it is difficult to add data to ACPI at runtime. gPXE creates an aBFT table for AoE and sBFT for SRP, and memdisk uses mBFT for MEMDISK at the moment.

It would be good to have some kind of common structure framework for these.

-hpa

#2 lore raw
Quoted 10 lines · H. Peter Anvin
Quoted 7 lines · Konrad Rzeszutek

What is iBFT?
The iBFT is an equivalent to the Boot Flag, except that its geared towards iSCSI and hence requires much more information (such as the IP of target, passwords, which device to login, etc). iBFT is a data structure populated by the BIOS or the NIC to contain this so that the OS can read it and login to the iSCSI and present the boot device to the initrd for mounting / FS.

I really don't see iBFT as equivalent to the boot flag at all. The boot

I think for somebody who might confuse iBFT with a bar of soap
the "equivalant" will put them in the right frame of mind. And yes it is not equivalant at all, should have said something like 'remotely akin' :-)

Quoted 5 lines

flag returns the status of the previous boot attempt; iBFT contains information about where to find the current root.

Unfortunately, we're increasingly seeing a proliferation of this kind of nonstandard ACPI tables, because it is difficult to add data to ACPI at

Keep in mind that iBFT is now a standard (woot!)

Quoted 2 lines

runtime. gPXE creates an aBFT table for AoE and sBFT for SRP, and memdisk uses mBFT for MEMDISK at the moment.

Oh man, didn't know those existed at all.

Quoted 2 lines

It would be good to have some kind of common structure framework for these.

I need to grok those tables some more to figure out what they all do.

#3 lore raw
Quoted 6 lines · Konrad Rzeszutek Wilk
Quoted 3 lines

Unfortunately, we're increasingly seeing a proliferation of this kind of nonstandard ACPI tables, because it is difficult to add data to ACPI at

Keep in mind that iBFT is now a standard (woot!)

Yes, but the discovery method is ad hoc, as opposed to the standard ACPI mechanisms.

Quoted 8 lines
Quoted 2 lines

runtime. gPXE creates an aBFT table for AoE and sBFT for SRP, and memdisk uses mBFT for MEMDISK at the moment.

Oh man, didn't know those existed at all.

Quoted 2 lines

It would be good to have some kind of common structure framework for these.

I need to grok those tables some more to figure out what they all do.

More or less the same thing as iBFT, but for AoE, SRP, or in-memory disk.

-hpa

#4 lore raw
Quoted 8 lines · On Monday 02 August 2010 16:26:59 H. Peter Anvin wrote:
Quoted 4 lines · Konrad Rzeszutek Wilk
Quoted 2 lines

Unfortunately, we're increasingly seeing a proliferation of this kind of nonstandard ACPI tables, because it is difficult to add data to ACPI at

Keep in mind that iBFT is now a standard (woot!)

Yes, but the discovery method is ad hoc, as opposed to the standard ACPI mechanisms.

The set of patches that I am asking Linus to pull now can use the ACPI table to find the table or in the fallback case the old method of scanning the memory. However if the machine has UEFI it will only do ACPI table lookup.

Can you point me what 'standard ACPI mechanism' is? Like sticking the code in the drivers/acpi ? And then having a generic driver to handle the [i,a,s,m]BFT tables and maybe some subordinate ones for specific pieces where the generic can't handle it?

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

It would be good to have some kind of common structure framework for these.

I need to grok those tables some more to figure out what they all do.

More or less the same thing as iBFT, but for AoE, SRP, or in-memory disk.

What is the tools state ? For iBFT, iscsi-initiator-utils scans the /sys/firmware directory to extract the relevant data and does its thing. Are there tools for AoE, SCSI RDMA Protocol (SRP), and in-memory disk?

#5 lore raw
Quoted 6 lines · Konrad Rzeszutek Wilk

Can you point me what 'standard ACPI mechanism' is? Like sticking the code in the drivers/acpi ? And then having a generic driver to handle the [i,a,s,m]BFT tables and maybe some subordinate ones for specific pieces where the generic can't handle it?

With the standard ACPI mechanism I meant RDSP -> {RSDT,XSDT} -> table. If it was easily possible to add SSDTs to this table structure then probably the best thing would have been to make them PnP devices.

Quoted 10 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

It would be good to have some kind of common structure framework for these.

I need to grok those tables some more to figure out what they all do.

More or less the same thing as iBFT, but for AoE, SRP, or in-memory disk.

What is the tools state ? For iBFT, iscsi-initiator-utils scans the /sys/firmware directory to extract the relevant data and does its thing. Are there tools for AoE, SCSI RDMA Protocol (SRP), and in-memory disk?

I don't know about AoE and SRP (Michael Brown <[email protected]> would know), but I have a tool which scans for mBFT directly out of /dev/mem, which is of course kind of ugly. It's shipped with the Syslinux distribution in the utils/ directory.

-hpa

Signature · 2 lines
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.
#6 lore raw
Quoted 9 lines
Quoted 4 lines

What is the tools state ? For iBFT, iscsi-initiator-utils scans the /sys/firmware directory to extract the relevant data and does its thing. Are there tools for AoE, SCSI RDMA Protocol (SRP), and in-memory disk?

I don't know about AoE and SRP (Michael Brown <[email protected]> would know), but I have a tool which scans for mBFT directly out of /dev/mem, which is of course kind of ugly. It's shipped with the Syslinux distribution in the utils/ directory.

Ok. Let me talk to Michael Brown about this and see what the interests are. This "unification" work would be post v2.6.36 thought.

#7 lore raw
In reply to H. Peter Anvin
Quoted 11 lines · H. Peter Anvin
Quoted 6 lines · Konrad Rzeszutek Wilk

Can you point me what 'standard ACPI mechanism' is? Like sticking the code in the drivers/acpi ? And then having a generic driver to handle the [i,a,s,m]BFT tables and maybe some subordinate ones for specific pieces where the generic can't handle it?

With the standard ACPI mechanism I meant RDSP -> {RSDT,XSDT} -> table. If it was easily possible to add SSDTs to this table structure then probably the best thing would have been to make them PnP devices.

The current iBFT specification and the current code (this push) includes the ability to specify the location of the iBFT table using the standard ACPI table mechanism, which is required on UEFI machines supporting iBFT, and which I would encourage non-UEFI machines to also use where possible. Obviously some systems, especially BIOS-based machines doing iSCSI boot on expansion cards, have difficulty with this, so we're not planning on de-supporting the old memory scanning method.

Signature · 6 lines
        Peter

Obviously, a major malfunction has occurred.
		-- Steve Nesbitt, voice of Mission Control, January 28, 1986

01234567890123456789012345678901234567890123456789012345678901234567890123456789
#8 lore raw
Quoted 9 lines · Peter Jones

The current iBFT specification and the current code (this push) includes the ability to specify the location of the iBFT table using the standard ACPI table mechanism, which is required on UEFI machines supporting iBFT, and which I would encourage non-UEFI machines to also use where possible. Obviously some systems, especially BIOS-based machines doing iSCSI boot on expansion cards, have difficulty with this, so we're not planning on de-supporting the old memory scanning method.

Not just expansion cards, but also post-INT19 software solutions.

-hpa

#9 lore raw

Re: [GIT PULL] iBFT features for v2.6.36 .. Question about LinuxCon.

Quoted 11 lines · H. Peter Anvin
Quoted 9 lines · Peter Jones

The current iBFT specification and the current code (this push) includes the ability to specify the location of the iBFT table using the standard ACPI table mechanism, which is required on UEFI machines supporting iBFT, and which I would encourage non-UEFI machines to also use where possible. Obviously some systems, especially BIOS-based machines doing iSCSI boot on expansion cards, have difficulty with this, so we're not planning on de-supporting the old memory scanning method.

Quoted 1 line

Not just expansion cards, but also post-INT19 software solutions.

Linus,

Aside the discussion we are having with hpa about expanding the iBFT or morphing it in another beast I realized that the LinuxCon is right around the corner.

I presume you are not going to be sitting in the back of a corner trying to merge all of the GIT pulls right before Aug 14/15th during the LinuxCon. So does that mean that the effective merge window is right before LinuxCon?

If that is the case, ping! Please pull this in before you pull in James's tree as he has patches (and post-post merge patches) that depend on this tree.

Or by any chance are you expanding the merge window by an extra week?

P.S.
Sorry to be so noise about this, but in the past I had missed two merge windows and I want to make sure I don't miss this one.

#10 lore raw

Re: [GIT PULL] iBFT features for v2.6.36 .. Question about LinuxCon.

Quoted 4 lines · Konrad Rzeszutek Wilk

If that is the case, ping! Please pull this in before you pull in James's tree as he has patches (and post-post merge patches) that depend
on this tree.

Actually, there's no dependency ... the pull I sent is the non-post-merge piece. It goes in independently of anyone else's tree. Post merge will wait on the resolution of the iBFT pull

James

#11 lore raw

Re: [GIT PULL] iBFT features for v2.6.36 .. Question about LinuxCon.

In reply to Konrad Rzeszutek Wilk
Quoted 5 lines · Konrad Rzeszutek Wilk

I presume you are not going to be sitting in the back of a corner trying to merge all of the GIT pulls right before Aug 14/15th during the LinuxCon. So does that mean that the effective merge window is right before LinuxCon?

I'm skipping LinuxCon in Boston, to instead do LinuxCon in Brazil two weeks later. So I will be at _a_ LinuxCon, but not the one you're thinking (quite frankly, Brazil is just more interesting than Boston - I've been to Boston before).

So I'm planning on sitting right at home, merging as usual. And LinuxCon is at the very end of the merge window, so I'd expect that any developer who sends me merge requests and is at LinuxCon will have sent his request before it.

Linus