GitLab

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

checksum and DQL (dynamic queue limits) are used by ixgbe driver. Add this to
liblcd as this may be used by other modules too in the future.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

A module parameter is added to the KLCD module, which is exposed by the kernel
in sysfs. The dispatch loop monitors for a change in this variable and sends a
message to LCD to exit, thereby paving way for LCD to gracefully exit.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

msleep is just a wrapper around udelay, which is a tight loop. msleep must be
replaced with a real timer in the future
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

A detailed list of functions supported
* register_netdev
* ether_setup
* eth_mac_addr
* eth_validate_addr
* free_netdev
* netif_carrier_off
* netif_carrier_on
* netif_device_attach
* netif_device_detach
* netif_set_real_num_rx_queues
* netif_set_real_num_tx_queues
* consume_skb
* unregister_netdev
* eth_platform_get_mac_address
* dev_addr_del
* device_set_wakeup_enable
* eth_get_headlen
* netif_tx_stop_all_queues
* pci_disable_pcie_error_reporting
* pci_bus_read_config_word
* pci_bus_write_config_word
* pci_cleanup_aer_uncorrect_error_status
* pci_disable_device
* pci_disable_msi
* pci_disable_msix
* pci_enable_msix_range
* pci_enable_msi_range
* pci_enable_pcie_error_reporting
* pcie_capability_read_word
* pcie_get_minimum_link
* pci_enable_device_mem
* pci_request_selected_regions
* pci_request_selected_regions_exclusive
* pci_set_master
* pci_save_state
* pci_restore_state
* pci_prepare_to_sleep
* pci_set_power_state
* pci_release_selected_regions
* pci_select_bars
* pci_wake_from_d3
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Replace
* dev_{warn,err,info,crit} with LIBLCD variants.
* v{m,z}alloc with kmalloc for now as vmalloc is not fully supported in the liblcd
implementation.
* {alloc,free}_pages with lcd variants

Mask functions that won't be used with LCD driver (for instance irq related
stuff). Add stubs for other functions that are unlikely to be RPC-ed or pulled
in.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Add the original driver, stripping all the files that are not necessary for our
82599 hardware.  Also remove (#ifdef) unneeded fancy features such as SRIOV,
PTP, HWMON, DEBUGFS, DCA, DCB. Compiles fine. There are unresolved symbols
though.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

PCI bar regions are accessed by corresponding drivers. ixgbe accesses only BAR0
region. Map this into the lcd using lcd_ioremap.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Instead of using naive schedule() that hogs the cpu, use a wait queue and wake
up the thread when someone pulls the plug.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Add support for remove function pointer and alloc_etherdev_mqs rpc. Also add
appropriate support in the cap for insertion/removal of newly introduced
containers.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

struct net_device is wrapped inside a container for the sake of LCDs. To
facilitate this, allocate memory for net_device_container instead of
net_device. Alter the api for accessing private area of net_device struct
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Populate boot, lcd, klcd modules of ixgbe and add a minimal ixgbe_main.c file
which tries to do pci_register_driver during load and unregister during exit.
Functions supported:
  * probe
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Update build scripts to build ixgbe test_module as well
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

oh! no. not again!
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

This script creates a tree structure with the necessary files for an LCD
module.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

   -- Needed to move ASYNC macro inside loop in pmfs dispatch loop
   -- Add pre_hook.h and post_hook.h headers to pmfs source files
   -- Fix get_current to return current_task global in liblcd/resolve.c,
      rather than use %gs percpu variable (was causing page fault)

Switching gears for a moment to fix some things in libasync.

ioremap_cache and friends just assume the memory is already
mapped in the address spaces, and does a phys -> virt translation.

request/release memory regions is a no-op inside the LCD. The
non-isolated glue for mount will need to invoke these functions.

unregister fs does a synchronize_rcu, which basically waits
for all cpu's to enter a "quiescent" state. The PMFS LCD thread
was never yielding as it waited for the response, so we had a
deadlock. I just needed to move cond_resched into the right place
in microkernel/run.c.

Kmalloc memory is not executable by default.

I also needed to make the maximum size ram map allocation bigger
in order to map 2 GBs.

We need a way to temporarily prevent external interrupts from
causing an LCD exit.

You have to be careful where you put your asyncs, especially
with poll recv. If you wrap the call to poll recv in an async,
you will consume lots of memory. I moved the async so it surrounds
the dispatch function only.

In the process, to debug, I updated libasync with THCDumpAllStacks
(I was trying to see what all the stacks were - they were all coming
from poll recv).

NOTE: While debugging, I tweaked the microkernel dump stack code
to work for THCDumpAllStacks (continue walking until we hit a
frame or return address). If you don't make this change,
THCDumpAllStacks will not crash, but it won't be informative either
(the stack walk will stop too soon).

I'm able to run at least 100 iterations now of mount/unmount without
any problems.

I wasn't acknowledging receipt of a couple responses, and the
async ipc buffer (the vfs tx buffer) slowly filled up after a
couple of test iterations.

But not all the way. It should be noted that if the async ipc
buffers are not big enough, you will get a deadlock. This is
because so many requests stay active before they are finally
ack'd with a response. This is due to the criss-crossing.

Needed to null out cnode metadata.

Wasn't storing metadata in cnode in cspace in a certain case.

I forgot I was memset'ing it to 0 in slab.c (and it wasn't necessary).

Command line string parsing functions are sucked in.

kthread spawning and waitqueue functions are no-op'd for now (these
are for pmfs's log cleaner). In the future, could redef these to
use libasync.

struct cptr_cache in libcap has struct mutex as a member. In its original
definition struct mutex has ifdef'd members, depending on the config chosen.
Since, liblcd build undef's certain config's explicitly (such as CONFIG_SMP),
this alters the size of struct mutex inside liblcd. To make the struct mutexes
equal in size, padding was introduced.

with 4.8, there is no difference in struct mutex size inside liblcd. So, make
this padding to zero for proper functioning of libcap.
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>

__get_cpu_var is replaced by this_cpu_read.

From commit,
1e02ce4c x86: Store a per-cpu shadow copy of CR4
CR4 reads take several cycles. So, store a per-cpu shadow copy of CR4.

So, replace direct reads with shadow_read()
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>