Skip to content

Conversation

frasersdev
Copy link

Following on from #1295

This Pull request enables the SPI 1 & 2 auxiliaries to compatible raspberry PIs

  • Backports the Auxiliary clock from 4.5 to 4.4
  • Minor config of the driver build environment to ensure they get built for CONFIG_ARCH_BCM2708 & CONFIG_ARCH_BCM2709 devices.
  • Adds the aux spi driver into the defconfigs as a module.
  • Adds the auxiliary and spi1/2 devices into the device tree in a disabled state
  • Provides device tree overlays which enables the devices and gives users a degree of control over how they are setup.

This pull request resolves issue #1034

These will be used for enabling UART1, SPI1, and SPI2.

Signed-off-by: Eric Anholt <[email protected]>
Acked-by: Rob Herring <[email protected]>
Signed-off-by: Michael Turquette <[email protected]>
There are a pair of SPI masters and a mini UART that were last minute
additions.  As a result, they didn't get integrated in the same way as
the other gates off of the VPU clock in CPRMAN.

Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Michael Turquette <[email protected]>

updated Makefile to preserve the rasoberry pi architectures
@frasersdev
Copy link
Author

one sec, there is a typo

@frasersdev
Copy link
Author

One thing that changed from the example I did was the (re)naming of the client enable (CE) lines (instead of client select CS). This is because the broadcom peripheral document uses 'CE' and that is what most people refer to them as on the PI. Its printed on scores of boards and headers etc. What do you think?

@pelwell
Copy link
Contributor

pelwell commented Feb 23, 2016

I'm not keen on that aspect of the change - the README doesn't use CE consistently which tells me that you aren't convinced either. The DT bindings for SPI refer to chipselects and use the abbreviation cs, so I think it is less confusing to keep that nomenclature.

By the way, I've made a few cosmetic tweaks to the third patch - you can find an updated version https://gist.github.com/pelwell/972aab58a718081502be, and if you are happy with my changes then you could update this PR with it and any CE->CS changes before I merge it.

Adds aux spi 1 & 2 devices to compatible raspberry PIs.
* Minor config of the driver build environment to ensure they get built
for CONFIG_ARCH_BCM2708 & CONFIG_ARCH_BCM2709 devices.
* Adds the aux spi driver into the defconfigs as a module.
* Adds the auxiliary and spi1/2 devices into the device tree in a
disabled state
* Provides decide tree overlays which enables the devices and gives
users a degree of control over how they are setup.
@frasersdev
Copy link
Author

ok, commit updated.

  • reverted ce->cs
  • aux patch included (tested & working fine BTW)

no other changes!

I'm happier with my README in this attempt.
I guess what was bothering me was there needed to be some reference back the the broadcom nomenclature as a reader may not make the connection that CS=CE (e.g. non-english speaker) - I think (hope) the reference aside the default numbering is enough without complicating everything.

@pelwell
Copy link
Contributor

pelwell commented Feb 24, 2016

It's a shame you edited your originals rather than my updated versions, but I've reapplied my changes and pushed the result.

@pelwell pelwell closed this Feb 24, 2016
@AlexSudnikov
Copy link

Hey everyone ! Sorry for a dumb question but...how do i enable theese aux SPI's. I mean , i only get spidev0.0 and spidev0.1 ( once a proper dtparam is added to the config ).
But if try to activate spi1 or spi2 ( following DTB's symbols names ) , like this:
dtparam=spi1=on
dtparam=spi2=on
Nothing happens...maybe someone could point out what is the problem , thx ! )
Btw, i'm using Rpi-3 model B , Linux 4.9.41-v7+

@pelwell
Copy link
Contributor

pelwell commented Sep 4, 2017

If you run dtoverlay -a | grep spi in a Raspbian shell you'll get:

  enc28j60-spi2
  raspidac3
  sc16is752-spi1
  spi-gpio35-39
  spi-rtc
  spi0-cs
  spi0-hw-cs
  spi1-1cs
  spi1-2cs
  spi1-3cs
  spi2-1cs
  spi2-2cs
  spi2-3cs

The overlays named spi1-* and spi2-*enable SPI1 and SPI2 with different numbers of chip-select/chip-enable lines.

$ dtoverlay -h spi1-2cs
Name:   spi1-2cs

Info:   Enables spi1 with two chip select (CS) lines and associated spidev
        dev nodes. The gpio pin numbers for the CS lines and spidev device node
        creation are configurable.
        N.B.: spi1 is only accessible on devices with a 40pin header, eg:
              A+, B+, Zero and PI2 B; as well as the Compute Module.

Usage:  dtoverlay=spi1-2cs,<param>=<val>

Params: cs0_pin                 GPIO pin for CS0 (default 18 - BCM SPI1_CE0).
        cs1_pin                 GPIO pin for CS1 (default 17 - BCM SPI1_CE1).
        cs0_spidev              Set to 'disabled' to stop the creation of a
                                userspace device node /dev/spidev1.0 (default
                                is 'okay' or enabled).
        cs1_spidev              Set to 'disabled' to stop the creation of a
                                userspace device node /dev/spidev1.1 (default
                                is 'okay' or enabled).

Note:

  1. Although we found the overlays and their usage from a shell, the above usage steps are meant to go in config.txt.
  2. Each interface is only available on a small number of pin groups - SPI1 is on GPIOs 16-21, while SPI2 is on GPIOs 40-49. Since only GPIOs 0-27 are brought out to the 40-pin header, SPI2 is only usable on a Compute Module.

@AlexSudnikov
Copy link

AlexSudnikov commented Sep 4, 2017

Ah, i see. Thanks for a reply pelwell !)
The only thing that is still a bit unclear to me is...how do i communicate with these aux. devices , i mean there is a python spidev module , which is used "for interfacing with SPI devices from user space via the spidev linux kernel driver " , however one still has to specify the exact device ( /dev/spidev0.0 for instance )
to communicate with. So the question is how does one actually use these chip select lines...but i think it's all about API :) Anyways, thx !

popcornmix pushed a commit that referenced this pull request Nov 3, 2020
When enabling qgroups we walk the tree_root and then add a qgroup item
for every root that we have.  This creates a lock dependency on the
tree_root and qgroup_root, which results in the following lockdep splat
(with tree locks using rwsem), eg. in tests btrfs/017 or btrfs/022:

  ======================================================
  WARNING: possible circular locking dependency detected
  5.9.0-default+ #1299 Not tainted
  ------------------------------------------------------
  btrfs/24552 is trying to acquire lock:
  ffff9142dfc5f630 (btrfs-quota-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]

  but task is already holding lock:
  ffff9142dfc5d0b0 (btrfs-root-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #1 (btrfs-root-00){++++}-{3:3}:
	 __lock_acquire+0x3fb/0x730
	 lock_acquire.part.0+0x6a/0x130
	 down_read_nested+0x46/0x130
	 __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
	 __btrfs_read_lock_root_node+0x3a/0x50 [btrfs]
	 btrfs_search_slot_get_root+0x11d/0x290 [btrfs]
	 btrfs_search_slot+0xc3/0x9f0 [btrfs]
	 btrfs_insert_item+0x6e/0x140 [btrfs]
	 btrfs_create_tree+0x1cb/0x240 [btrfs]
	 btrfs_quota_enable+0xcd/0x790 [btrfs]
	 btrfs_ioctl_quota_ctl+0xc9/0xe0 [btrfs]
	 __x64_sys_ioctl+0x83/0xa0
	 do_syscall_64+0x2d/0x70
	 entry_SYSCALL_64_after_hwframe+0x44/0xa9

  -> #0 (btrfs-quota-00){++++}-{3:3}:
	 check_prev_add+0x91/0xc30
	 validate_chain+0x491/0x750
	 __lock_acquire+0x3fb/0x730
	 lock_acquire.part.0+0x6a/0x130
	 down_read_nested+0x46/0x130
	 __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
	 __btrfs_read_lock_root_node+0x3a/0x50 [btrfs]
	 btrfs_search_slot_get_root+0x11d/0x290 [btrfs]
	 btrfs_search_slot+0xc3/0x9f0 [btrfs]
	 btrfs_insert_empty_items+0x58/0xa0 [btrfs]
	 add_qgroup_item.part.0+0x72/0x210 [btrfs]
	 btrfs_quota_enable+0x3bb/0x790 [btrfs]
	 btrfs_ioctl_quota_ctl+0xc9/0xe0 [btrfs]
	 __x64_sys_ioctl+0x83/0xa0
	 do_syscall_64+0x2d/0x70
	 entry_SYSCALL_64_after_hwframe+0x44/0xa9

  other info that might help us debug this:

   Possible unsafe locking scenario:

	 CPU0                    CPU1
	 ----                    ----
    lock(btrfs-root-00);
				 lock(btrfs-quota-00);
				 lock(btrfs-root-00);
    lock(btrfs-quota-00);

   *** DEADLOCK ***

  5 locks held by btrfs/24552:
   #0: ffff9142df431478 (sb_writers#10){.+.+}-{0:0}, at: mnt_want_write_file+0x22/0xa0
   #1: ffff9142f9b10cc0 (&fs_info->subvol_sem){++++}-{3:3}, at: btrfs_ioctl_quota_ctl+0x7b/0xe0 [btrfs]
   #2: ffff9142f9b11a08 (&fs_info->qgroup_ioctl_lock){+.+.}-{3:3}, at: btrfs_quota_enable+0x3b/0x790 [btrfs]
   #3: ffff9142df431698 (sb_internal#2){.+.+}-{0:0}, at: start_transaction+0x406/0x510 [btrfs]
   #4: ffff9142dfc5d0b0 (btrfs-root-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]

  stack backtrace:
  CPU: 1 PID: 24552 Comm: btrfs Not tainted 5.9.0-default+ #1299
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-59-gc9ba527-rebuilt.opensuse.org 04/01/2014
  Call Trace:
   dump_stack+0x77/0x97
   check_noncircular+0xf3/0x110
   check_prev_add+0x91/0xc30
   validate_chain+0x491/0x750
   __lock_acquire+0x3fb/0x730
   lock_acquire.part.0+0x6a/0x130
   ? __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   ? lock_acquire+0xc4/0x140
   ? __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   down_read_nested+0x46/0x130
   ? __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   ? btrfs_root_node+0xd9/0x200 [btrfs]
   __btrfs_read_lock_root_node+0x3a/0x50 [btrfs]
   btrfs_search_slot_get_root+0x11d/0x290 [btrfs]
   btrfs_search_slot+0xc3/0x9f0 [btrfs]
   btrfs_insert_empty_items+0x58/0xa0 [btrfs]
   add_qgroup_item.part.0+0x72/0x210 [btrfs]
   btrfs_quota_enable+0x3bb/0x790 [btrfs]
   btrfs_ioctl_quota_ctl+0xc9/0xe0 [btrfs]
   __x64_sys_ioctl+0x83/0xa0
   do_syscall_64+0x2d/0x70
   entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fix this by dropping the path whenever we find a root item, add the
qgroup item, and then re-lookup the root item we found and continue
processing roots.

Reported-by: David Sterba <[email protected]>
Reviewed-by: Filipe Manana <[email protected]>
Signed-off-by: Josef Bacik <[email protected]>
Reviewed-by: David Sterba <[email protected]>
Signed-off-by: David Sterba <[email protected]>
popcornmix pushed a commit that referenced this pull request Nov 17, 2020
[ Upstream commit 5223cc6 ]

When enabling qgroups we walk the tree_root and then add a qgroup item
for every root that we have.  This creates a lock dependency on the
tree_root and qgroup_root, which results in the following lockdep splat
(with tree locks using rwsem), eg. in tests btrfs/017 or btrfs/022:

  ======================================================
  WARNING: possible circular locking dependency detected
  5.9.0-default+ #1299 Not tainted
  ------------------------------------------------------
  btrfs/24552 is trying to acquire lock:
  ffff9142dfc5f630 (btrfs-quota-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]

  but task is already holding lock:
  ffff9142dfc5d0b0 (btrfs-root-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #1 (btrfs-root-00){++++}-{3:3}:
	 __lock_acquire+0x3fb/0x730
	 lock_acquire.part.0+0x6a/0x130
	 down_read_nested+0x46/0x130
	 __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
	 __btrfs_read_lock_root_node+0x3a/0x50 [btrfs]
	 btrfs_search_slot_get_root+0x11d/0x290 [btrfs]
	 btrfs_search_slot+0xc3/0x9f0 [btrfs]
	 btrfs_insert_item+0x6e/0x140 [btrfs]
	 btrfs_create_tree+0x1cb/0x240 [btrfs]
	 btrfs_quota_enable+0xcd/0x790 [btrfs]
	 btrfs_ioctl_quota_ctl+0xc9/0xe0 [btrfs]
	 __x64_sys_ioctl+0x83/0xa0
	 do_syscall_64+0x2d/0x70
	 entry_SYSCALL_64_after_hwframe+0x44/0xa9

  -> #0 (btrfs-quota-00){++++}-{3:3}:
	 check_prev_add+0x91/0xc30
	 validate_chain+0x491/0x750
	 __lock_acquire+0x3fb/0x730
	 lock_acquire.part.0+0x6a/0x130
	 down_read_nested+0x46/0x130
	 __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
	 __btrfs_read_lock_root_node+0x3a/0x50 [btrfs]
	 btrfs_search_slot_get_root+0x11d/0x290 [btrfs]
	 btrfs_search_slot+0xc3/0x9f0 [btrfs]
	 btrfs_insert_empty_items+0x58/0xa0 [btrfs]
	 add_qgroup_item.part.0+0x72/0x210 [btrfs]
	 btrfs_quota_enable+0x3bb/0x790 [btrfs]
	 btrfs_ioctl_quota_ctl+0xc9/0xe0 [btrfs]
	 __x64_sys_ioctl+0x83/0xa0
	 do_syscall_64+0x2d/0x70
	 entry_SYSCALL_64_after_hwframe+0x44/0xa9

  other info that might help us debug this:

   Possible unsafe locking scenario:

	 CPU0                    CPU1
	 ----                    ----
    lock(btrfs-root-00);
				 lock(btrfs-quota-00);
				 lock(btrfs-root-00);
    lock(btrfs-quota-00);

   *** DEADLOCK ***

  5 locks held by btrfs/24552:
   #0: ffff9142df431478 (sb_writers#10){.+.+}-{0:0}, at: mnt_want_write_file+0x22/0xa0
   #1: ffff9142f9b10cc0 (&fs_info->subvol_sem){++++}-{3:3}, at: btrfs_ioctl_quota_ctl+0x7b/0xe0 [btrfs]
   #2: ffff9142f9b11a08 (&fs_info->qgroup_ioctl_lock){+.+.}-{3:3}, at: btrfs_quota_enable+0x3b/0x790 [btrfs]
   #3: ffff9142df431698 (sb_internal#2){.+.+}-{0:0}, at: start_transaction+0x406/0x510 [btrfs]
   #4: ffff9142dfc5d0b0 (btrfs-root-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]

  stack backtrace:
  CPU: 1 PID: 24552 Comm: btrfs Not tainted 5.9.0-default+ #1299
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-59-gc9ba527-rebuilt.opensuse.org 04/01/2014
  Call Trace:
   dump_stack+0x77/0x97
   check_noncircular+0xf3/0x110
   check_prev_add+0x91/0xc30
   validate_chain+0x491/0x750
   __lock_acquire+0x3fb/0x730
   lock_acquire.part.0+0x6a/0x130
   ? __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   ? lock_acquire+0xc4/0x140
   ? __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   down_read_nested+0x46/0x130
   ? __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   __btrfs_tree_read_lock+0x35/0x1c0 [btrfs]
   ? btrfs_root_node+0xd9/0x200 [btrfs]
   __btrfs_read_lock_root_node+0x3a/0x50 [btrfs]
   btrfs_search_slot_get_root+0x11d/0x290 [btrfs]
   btrfs_search_slot+0xc3/0x9f0 [btrfs]
   btrfs_insert_empty_items+0x58/0xa0 [btrfs]
   add_qgroup_item.part.0+0x72/0x210 [btrfs]
   btrfs_quota_enable+0x3bb/0x790 [btrfs]
   btrfs_ioctl_quota_ctl+0xc9/0xe0 [btrfs]
   __x64_sys_ioctl+0x83/0xa0
   do_syscall_64+0x2d/0x70
   entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fix this by dropping the path whenever we find a root item, add the
qgroup item, and then re-lookup the root item we found and continue
processing roots.

Reported-by: David Sterba <[email protected]>
Reviewed-by: Filipe Manana <[email protected]>
Signed-off-by: Josef Bacik <[email protected]>
Reviewed-by: David Sterba <[email protected]>
Signed-off-by: David Sterba <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants