layer, but with a twist.
The twist has to do with the fact that Microsoft supports structured
exception handling in kernel mode. On the i386 arch, exception handling
is implemented by hanging an exception registration list off the
Thread Environment Block (TEB), and the TEB is accessed via the %fs
register. The problem is, we use %fs as a pointer to the pcpu stucture,
which means any driver that tries to write through %fs:0 will overwrite
the curthread pointer and make a serious mess of things.
To get around this, Project Evil now creates a special entry in
the GDT on each processor. When we call into Windows code, a context
switch routine will fix up %fs so it points to our new descriptor,
which in turn points to a fake TEB. When the Windows code returns,
or calls out to an external routine, we swap %fs back again. Currently,
Project Evil makes use of GDT slot 7, which is all 0s by default.
I fully expect someone to jump up and say I can't do that, but I
couldn't find any code that makes use of this entry anywhere. Sadly,
this was the only method I could come up with that worked on both
UP and SMP. (Modifying the LDT works on UP, but becomes incredibly
complicated on SMP.) If necessary, the context switching stuff can
be yanked out while preserving the convention calling wrappers.
(Fortunately, it looks like Microsoft uses some special epilog/prolog
code on amd64 to implement exception handling, so the same nastiness
won't be necessary on that arch.)
The advantages are:
- Any driver that uses %fs as though it were a TEB pointer won't
clobber pcpu.
- All the __stdcall/__fastcall/__regparm stuff that's specific to
gcc goes away.
Also, while I'm here, switch NdisGetSystemUpTime() back to using
nanouptime() again. It turns out nanouptime() is way more accurate
than just using ticks(). On slower machines, the Atheros drivers
I tested seem to take a long time to associate due to the loss
in accuracy.
work on SMP" saga. After several weeks and much gnashing of teeth,
I have finally tracked down all the problems, despite their best
efforts to confound and annoy me.
Problem nunmber one: the Atheros windows driver is _NOT_ a de-serialized
miniport! It used to be that NDIS drivers relied on the NDIS library
itself for all their locking and serialization needs. Transmit packet
queues were all handled internally by NDIS, and all calls to
MiniportXXX() routines were guaranteed to be appropriately serialized.
This proved to be a performance problem however, and Microsoft
introduced de-serialized miniports with the NDIS 5.x spec. Microsoft
still supports serialized miniports, but recommends that all new drivers
written for Windows XP and later be deserialized. Apparently Atheros
wasn't listening when they said this.
This means (among other things) that we have to serialize calls to
MiniportSendPackets(). We also have to serialize calls to MiniportTimer()
that are triggered via the NdisMInitializeTimer() routine. It finally
dawned on me why NdisMInitializeTimer() takes a special
NDIS_MINIPORT_TIMER structure and a pointer to the miniport block:
the timer callback must be serialized, and it's only by saving the
miniport block handle that we can get access to the serialization
lock during the timer callback.
Problem number two: haunted hardware. The thing that was _really_
driving me absolutely bonkers for the longest time is that, for some
reason I couldn't understand, my test machine would occasionally freeze
or more frustratingly, reset completely. That's reset and in *pow!*
back to the BIOS startup. No panic, no crashdump, just a reset. This
appeared to happen most often when MiniportReset() was called. (As
to why MiniportReset() was being called, see problem three below.)
I thought maybe I had created some sort of horrible deadlock
condition in the process of adding the serialization, but after three
weeks, at least 6 different locking implementations and heroic efforts
to debug the spinlock code, the machine still kept resetting. Finally,
I started single stepping through the MiniportReset() routine in
the driver using the kernel debugger, and this ultimately led me to
the source of the problem.
One of the last things the Atheros MiniportReset() routine does is
call NdisReadPciSlotInformation() several times to inspect a portion
of the device's PCI config space. It reads the same chunk of config
space repeatedly, in rapid succession. Presumeably, it's polling
the hardware for some sort of event. The reset occurs partway through
this process. I discovered that when I single-stepped through this
portion of the routine, the reset didn't occur. So I inserted a 1
microsecond delay into the read loop in NdisReadPciSlotInformation().
Suddenly, the reset was gone!!
I'm still very puzzled by the whole thing. What I suspect is happening
is that reading the PCI config space so quickly is causing a severe
PCI bus error. My test system is a Sun w2100z dual Opteron system,
and the NIC is a miniPCI card mounted in a miniPCI-to-PCI carrier card,
plugged into a 100Mhz PCI slot. It's possible that this combination of
hardware causes a bus protocol violation in this scenario which leads
to a fatal machine check. This is pure speculation though. Really all I
know for sure is that inserting the delay makes the problem go away.
(To quote Homer Simpson: "I don't know how it works, but fire makes
it good!")
Problem number three: NdisAllocatePacket() needs to make sure to
initialize the npp_validcounts field in the 'private' section of
the NDIS_PACKET structure. The reason if_ndis was calling the
MiniportReset() routine in the first place is that packet transmits
were sometimes hanging. When sending a packet, an NDIS driver will
call NdisQueryPacket() to learn how many physical buffers the packet
resides in. NdisQueryPacket() is actually a macro, which traverses
the NDIS_BUFFER list attached to the NDIS_PACKET and stashes some
of the results in the 'private' section of the NDIS_PACKET. It also
sets the npp_validcounts field to TRUE To indicate that the results are
now valid. The problem is, now that if_ndis creates a pool of transmit
packets via NdisAllocatePacketPool(), it's important that each time
a new packet is allocated via NdisAllocatePacket() that validcounts
be initialized to FALSE. If it isn't, and a previously transmitted
NDIS_PACKET is pulled out of the pool, it may contain stale data
from a previous transmission which won't get updated by NdisQueryPacket().
This would cause the driver to miscompute the number of fragments
for a given packet, and botch the transmission.
Fixing these three problems seems to make the Atheros driver happy
on SMP, which hopefully means other serialized miniports will be
happy too.
And there was much rejoicing.
Other stuff fixed along the way:
- Modified ndis_thsuspend() to take a mutex as an argument. This
allows KeWaitForSingleObject() and KeWaitForMultipleObjects() to
avoid any possible race conditions with other routines that
use the dispatcher lock.
- Fixed KeCancelTimer() so that it returns the correct value for
'pending' according to the Microsoft documentation
- Modfied NdisGetSystemUpTime() to use ticks and hz rather than
calling nanouptime(). Also added comment that this routine wraps
after 49.7 days.
- Added macros for KeAcquireSpinLock()/KeReleaseSpinLock() to hide
all the MSCALL() goop.
- For x86, KeAcquireSpinLockRaiseToDpc() needs to be a separate
function. This is because it's supposed to be _stdcall on the x86
arch, whereas KeAcquireSpinLock() is supposed to be _fastcall.
On amd64, all routines use the same calling convention so we can
just map KeAcquireSpinLockRaiseToDpc() directly to KfAcquireSpinLock()
and it will work. (The _fastcall attribute is a no-op on amd64.)
- Implement and use IoInitializeDpcRequest() and IoRequestDpc() (they're
just macros) and use them for interrupt handling. This allows us to
move the ndis_intrtask() routine from if_ndis.c to kern_ndis.c.
- Fix the MmInitializeMdl() macro so that is uses sizeof(vm_offset_t)
when computing mdl_size instead of uint32_t, so that it matches the
MmSizeOfMdl() routine.
- Change a could of M_WAITOKs to M_NOWAITs in the unicode routines in
subr_ndis.c.
- Use the dispatcher lock a little more consistently in subr_ntoskrnl.c.
- Get rid of the "wait for link event" hack in ndis_init(). Now that
I fixed NdisReadPciSlotInformation(), it seems I don't need it anymore.
This should fix the witness panic a couple of people have reported.
- Use MSCALL1() when calling the MiniportHangCheck() function in
ndis_ticktask(). I accidentally missed this one when adding the
wrapping for amd64.
at some point result in a status event being triggered (it should
be a link down event: the Microsoft driver design guide says you
should generate one when the NIC is initialized). Some drivers
generate the event during MiniportInitialize(), such that by the
time MiniportInitialize() completes, the NIC is ready to go. But
some drivers, in particular the ones for Atheros wireless NICs,
don't generate the event until after a device interrupt occurs
at some point after MiniportInitialize() has completed.
The gotcha is that you have to wait until the link status event
occurs one way or the other before you try to fiddle with any
settings (ssid, channel, etc...). For the drivers that set the
event sycnhronously this isn't a problem, but for the others
we have to pause after calling ndis_init_nic() and wait for the event
to arrive before continuing. Failing to wait can cause big trouble:
on my SMP system, calling ndis_setstate_80211() after ndis_init_nic()
completes, but _before_ the link event arrives, will lock up or
reset the system.
What we do now is check to see if a link event arrived while
ndis_init_nic() was running, and if it didn't we msleep() until
it does.
Along the way, I discovered a few other problems:
- Defered procedure calls run at PASSIVE_LEVEL, not DISPATCH_LEVEL.
ntoskrnl_run_dpc() has been fixed accordingly. (I read the documentation
wrong.)
- Similarly, the NDIS interrupt handler, which is essentially a
DPC, also doesn't need to run at DISPATCH_LEVEL. ndis_intrtask()
has been fixed accordingly.
- MiniportQueryInformation() and MiniportSetInformation() run at
DISPATCH_LEVEL, and each request must complete before another
can be submitted. ndis_get_info() and ndis_set_info() have been
fixed accordingly.
- Turned the sleep lock that guards the NDIS thread job list into
a spin lock. We never do anything with this lock held except manage
the job list (no other locks are held), so it's safe to do this,
and it's possible that ndis_sched() and ndis_unsched() can be
called from DISPATCH_LEVEL, so using a sleep lock here is
semantically incorrect. Also updated subr_witness.c to add the
lock to the order list.
for a given device in some circumstances, so move the PDO creation
to the attach routine so we don't end up creating two PDOs.
Also, when we skip the call to ndis_convert_res() in if_ndis.c:ndis_attach(),
initialize sc->ndis_block->nmb_rlist to NULL. We don't explicitly zero
the miniport block, so this will make sure ndis_unload_driver() does
the right thing.
when we create a PDO, the driver_object associated with it is that
of the parent driver, not the driver we're trying to attach. For
example, if we attach a PCI device, the PDO we pass to the NdisAddDevice()
function should contain a pointer to fake_pci_driver, not to the NDIS
driver itself. For PCI or PCMCIA devices this doesn't matter because
the child never needs to talk to the parent bus driver, but for USB,
the child needs to be able to send IRPs to the parent USB bus driver, and
for that to work the parent USB bus driver has to be hung off the PDO.
This involves modifying windrv_lookup() so that we can search for
bus drivers by name, if necessary. Our fake bus drivers attach themselves
as "PCI Bus," "PCCARD Bus" and "USB Bus," so we can search for them
using those names.
The individual attachment stubs now create and attach PDOs to the
parent bus drivers instead of hanging them off the NDIS driver's
object, and in if_ndis.c, we now search for the correct driver
object depending on the bus type, and use that to find the correct PDO.
With this fix, I can get my sample USB ethernet driver to deliver
an IRP to my fake parent USB bus driver's dispatch routines.
- Add stub modules for USB support: subr_usbd.c, usbd_var.h and
if_ndis_usb.c. The subr_usbd.c module is hooked up the build
but currently doesn't do very much. It provides the stub USB
parent driver object and a dispatch routine for
IRM_MJ_INTERNAL_DEVICE_CONTROL. The only exported function at
the moment is USBD_GetUSBDIVersion(). The if_ndis_usb.c stub
compiles, but is not hooked up to the build yet. I'm putting
these here so I can keep them under source code control as I
flesh them out.
Ville-Pertti Keinonen (will at exomi dot comohmygodnospampleasekthx)
deserves a big thanks for submitting initial patches to make it
work. I have mangled his contributions appropriately.
The main gotcha with Windows/x86-64 is that Microsoft uses a different
calling convention than everyone else. The standard ABI requires using
6 registers for argument passing, with other arguments on the stack.
Microsoft uses only 4 registers, and requires the caller to leave room
on the stack for the register arguments incase the callee needs to
spill them. Unlike x86, where Microsoft uses a mix of _cdecl, _stdcall
and _fastcall, all routines on Windows/x86-64 uses the same convention.
This unfortunately means that all the functions we export to the
driver require an intermediate translation wrapper. Similarly, we have
to wrap all calls back into the driver binary itself.
The original patches provided macros to wrap every single routine at
compile time, providing a secondary jump table with a customized
wrapper for each exported routine. I decided to use a different approach:
the call wrapper for each function is created from a template at
runtime, and the routine to jump to is patched into the wrapper as
it is created. The subr_pe module has been modified to patch in the
wrapped function instead of the original. (On x86, the wrapping
routine is a no-op.)
There are some minor API differences that had to be accounted for:
- KeAcquireSpinLock() is a real function on amd64, not a macro wrapper
around KfAcquireSpinLock()
- NdisFreeBuffer() is actually IoFreeMdl(). I had to change the whole
NDIS_BUFFER API a bit to accomodate this.
Bugs fixed along the way:
- IoAllocateMdl() always returned NULL
- kern_windrv.c:windrv_unload() wasn't releasing private driver object
extensions correctly (found thanks to memguard)
This has only been tested with the driver for the Broadcom 802.11g
chipset, which was the only Windows/x86-64 driver I could find.
Windows DRIVER_OBJECT and DEVICE_OBJECT mechanism so that we can
simulate driver stacking.
In Windows, each loaded driver image is attached to a DRIVER_OBJECT
structure. Windows uses the registry to match up a given vendor/device
ID combination with a corresponding DRIVER_OBJECT. When a driver image
is first loaded, its DriverEntry() routine is invoked, which sets up
the AddDevice() function pointer in the DRIVER_OBJECT and creates
a dispatch table (based on IRP major codes). When a Windows bus driver
detects a new device, it creates a Physical Device Object (PDO) for
it. This is a DEVICE_OBJECT structure, with semantics analagous to
that of a device_t in FreeBSD. The Windows PNP manager will invoke
the driver's AddDevice() function and pass it pointers to the DRIVER_OBJECT
and the PDO.
The AddDevice() function then creates a new DRIVER_OBJECT structure of
its own. This is known as the Functional Device Object (FDO) and
corresponds roughly to a private softc instance. The driver uses
IoAttachDeviceToDeviceStack() to add this device object to the
driver stack for this PDO. Subsequent drivers (called filter drivers
in Windows-speak) can be loaded which add themselves to the stack.
When someone issues an IRP to a device, it travel along the stack
passing through several possible filter drivers until it reaches
the functional driver (which actually knows how to talk to the hardware)
at which point it will be completed. This is how Windows achieves
driver layering.
Project Evil now simulates most of this. if_ndis now has a modevent
handler which will use MOD_LOAD and MOD_UNLOAD events to drive the
creation and destruction of DRIVER_OBJECTs. (The load event also
does the relocation/dynalinking of the image.) We don't have a registry,
so the DRIVER_OBJECTS are stored in a linked list for now. Eventually,
the list entry will contain the vendor/device ID list extracted from
the .INF file. When ndis_probe() is called and detectes a supported
device, it will create a PDO for the device instance and attach it
to the DRIVER_OBJECT just as in Windows. ndis_attach() will then call
our NdisAddDevice() handler to create the FDO. The NDIS miniport block
is now a device extension hung off the FDO, just as it is in Windows.
The miniport characteristics table is now an extension hung off the
DRIVER_OBJECT as well (the characteristics are the same for all devices
handled by a given driver, so they don't need to be per-instance.)
We also do an IoAttachDeviceToDeviceStack() to put the FDO on the
stack for the PDO. There are a couple of fake bus drivers created
for the PCI and pccard buses. Eventually, there will be one for USB,
which will actually accept USB IRP.s
Things should still work just as before, only now we do things in
the proper order and maintain the correct framework to support passing
IRPs between drivers.
Various changes:
- corrected the comments about IRQL handling in subr_hal.c to more
accurately reflect reality
- update ndiscvt to make the drv_data symbol in ndis_driver_data.h a
global so that if_ndis_pci.o and/or if_ndis_pccard.o can see it.
- Obtain the softc pointer from the miniport block by referencing
the PDO rather than a private pointer of our own (nmb_ifp is no
longer used)
- implement IoAttachDeviceToDeviceStack(), IoDetachDevice(),
IoGetAttachedDevice(), IoAllocateDriverObjectExtension(),
IoGetDriverObjectExtension(), IoCreateDevice(), IoDeleteDevice(),
IoAllocateIrp(), IoReuseIrp(), IoMakeAssociatedIrp(), IoFreeIrp(),
IoInitializeIrp()
- fix a few mistakes in the driver_object and device_object definitions
- add a new module, kern_windrv.c, to handle the driver registration
and relocation/dynalinkign duties (which don't really belong in
kern_ndis.c).
- made ndis_block and ndis_chars in the ndis_softc stucture pointers
and modified all references to it
- fixed NdisMRegisterMiniport() and NdisInitializeWrapper() so they
work correctly with the new driver_object mechanism
- changed ndis_attach() to call NdisAddDevice() instead of ndis_load_driver()
(which is now deprecated)
- used ExAllocatePoolWithTag()/ExFreePool() in lookaside list routines
instead of kludged up alloc/free routines
- added kern_windrv.c to sys/modules/ndis/Makefile and files.i386.
USB device support):
- Convert all of my locally chosen function names to their actual
Windows equivalents, where applicable. This is a big no-op change
since it doesn't affect functionality, but it helps avoid a bit
of confusion (it's now a lot easier to see which functions are
emulated Windows API routines and which are just locally defined).
- Turn ndis_buffer into an mdl, like it should have been. The structure
is the same, but now it belongs to the subr_ntoskrnl module.
- Implement a bunch of MDL handling macros from Windows and use them where
applicable.
- Correct the implementation of IoFreeMdl().
- Properly implement IoAllocateMdl() and MmBuildMdlForNonPagedPool().
- Add the definitions for struct irp and struct driver_object.
- Add IMPORT_FUNC() and IMPORT_FUNC_MAP() macros to make formatting
the module function tables a little cleaner. (Should also help
with AMD64 support later on.)
- Fix if_ndis.c to use KeRaiseIrql() and KeLowerIrql() instead of
the previous calls to hal_raise_irql() and hal_lower_irql() which
have been renamed.
The function renaming generated a lot of churn here, but there should
be very little operational effect.
now, but it's possible for ndis_reset_nic() to sleep (sometimes the
MiniportReset() method returns NDIS_STATUS_PENDING and we have
to wait for completion). To get around this, execute the ndis_reset_nic()
routine in the NDIS_TASKQUEUE thread.
- In subr_ndis.c:ndis_allocate_sharemem(), create the busdma tags
used for shared memory allocations with a lowaddr of 0x3E7FFFFF.
This forces the buffers to be mapped to physical/bus addresses within
the first 1GB of physical memory. It seems that at least one card
(Linksys Instant Wireless PCI V2.7) depends on this behavior. I
don't know if this is a hardware restriction, or if the NDIS
driver for this card is truncating the addresses itself, but using
physical/bus addresses beyong the 1GB limit causes initialization
failures.
- Create am NDIS_INITIALIZED() macro in if_ndisvar.h and use it in
if_ndis.c to test whether the device has been initialized rather
than checking for the presence of the IFF_UP flag in if_flags.
While debugging the previous problem, I noticed that bringing
up the device would always produce failures from ndis_setmulti().
It turns out that the following steps now occur during device
initialization:
- IFF_UP flag is set in if_flags
- ifp->if_ioctl() called with SIOCSIFADDR (which we don't handle)
- ifp->if_ioctl() called with SIOCADDMULTI
- ifp->if_ioctl() called with SIOCADDMULTI (again)
- ifp->if_ioctl() called with SIOCADDMULTI (yet again)
- ifp->if_ioctl() called with SIOCSIFFLAGS
Setting the receive filter and multicast filters can only be done
when the underlying NDIS driver has been initialized, which is done
by ifp->if_init(). However, we don't call ifp->if_init() until
ifp->if_ioctl() is called with SIOCSIFFLAGS and IFF_UP has been
set. It appears that now, the network stack tries to add multicast
addresses to interface's filter before those steps occur. Normally,
ndis_setmulti() would trap this condition by checking for the IFF_UP
flag, but the network code has in fact set this flag already, so
ndis_setmulti() is fooled into thinking the interface has been
initialized when it really hasn't.
It turns out this is usually harmless because the ifp->if_init()
routine (in this case ndis_init()) will set up the multicast
filter when it initializes the hardware anyway, and the underlying
routines (ndis_get_info()/ndis_set_info()) know that the driver/NIC
haven't been initialized yet, but you end up spurious error messages
on the console all the time.
Something tells me this new behavior isn't really correct. I think
the intention was to fix it so that ifp->if_init() is only called
once when we ifconfig an interface up, but the end result seems a
little bogus: the change of the IFF_UP flag should be propagated
down to the driver before calling any other ioctl() that might actually
require the hardware to be up and running.
- In subr_ndis.c, my_strcasecmp() actually behaved like my_strncasecmp():
we really need it to behave like the former, not the latter. (It was
falsely matching "RadioEnable", which defaults to 1 with "RadioEnableHW"
which the driver creates itself and to 0, because we were using
strlen("RadioEnable") as the length to test. This caused the radio to
always be turned off. :( )
- In if_ndis.c, only set IEEE80211_CHAN_A for channels if we actually
set any IEEE80211_MODE_11A rates. (ieee80211_attach() will "helpfully"
add IEEE80211_MODE_11A to ic_modecaps for you if you initialize any
802.11a channels. This caused "ndis0: 11a rates:" to erroneously be
displayed during driver load.)
- Also in if_ndis.c, when using TESTSETRATE() to add in any missing 802.11b
rates, remember to OR the rates with IEEE80211_RATE_BASIC, otherwise
comparing against existing basic rates won't match. (1, 2, 5.5 and
11Mbps are basic rates, according to the 802.11b spec.) This erroneously
cause 11Mbps to be added to the 11b rate list twice.
make the key name matching case-insensitive. There are some drivers
and .inf files that have mismatched cases, e.g. the driver will look
for "AdhocBand" whereas the .inf file specifies a registry key to be
created called "AdHocBand." The mismatch is probably a typo that went
undetected (so much for QA), but since Windows seems to be case-insensitive,
we should be too.
In if_ndis.c, initialize rates and channels correctly so that specify
frequences correctly when trying to set channels in the 5Ghz band, and
so that 802.11b rates show up for some a/b/g cards (which otherwise
appear to have no 802.11b modes).
Also, when setting OID_802_11_CONFIGURATION in ndis_80211_setstate(),
provide default values for the beacon interval, ATIM window and dwelltime.
The Atheros "Aries" driver will crash if you try to select ad-hoc mode
and leave the beacon interval set to 0: it blindly uses this value and
does a division by 0 in the interrupt handler, causing an integer
divide trap.
- Use the dh_inserted member of the dispatch header in the Windows
timer structure to indicate that the timer has been "inserted into
the timer queue" (i.e. armed via timeout()). Use this as the value
to return to the caller in KeCancelTimer(). Previously, I was using
callout_pending(), but you can't use that with timeout()/untimeout()
without creating a potential race condition.
- Make ntoskrnl_init_timer() just a wrapper around ntoskrnl_init_timer_ex()
(reduces some code duplication).
- Drop Giant when entering if_ndis.c:ndis_tick() and
subr_ntorkrnl.c:ntoskrnl_timercall(). At the moment, I'm forced to
use system callwheel via timeout()/untimeout() to handle timers rather
than the callout API (struct callout is too big to fit inside the
Windows struct KTIMER, so I'm kind of hosed). Unfortunately, all
the callouts in the callwhere are not marked as MPSAFE, so when
one of them fires, it implicitly acquires Giant before invoking the
callback routine (and releases it when it returns). I don't need to
hold Giant, but there's no way to stop the callout code from acquiring
it as long as I'm using timeout()/untimeout(), so for now we cheat
by just dropping Giant right away (and re-acquiring it right before
the routine returns so keep the callout code happy). At some point,
I will need to solve this better, but for now this should be a suitable
workaround.
(I hope.)
My original instinct to make ndis_return_packet() asynchronous was correct.
Making ndis_rxeof() submit packets to the stack asynchronously fixes
one recursive spinlock acquisition, but it's also possible for it to
happen via the ndis_txeof() path too. So:
- In if_ndis.c, revert ndis_rxeof() to its old behavior (and don't bother
putting ndis_rxeof_serial() back since we don't need it anymore).
- In kern_ndis.c, make ndis_return_packet() submit the call to the
MiniportReturnPacket() function to the "ndis swi" thread so that
it always happens in another context no matter who calls it.
wireless ever since I added the new spinlock code. Previously, I added
a special ndis_rxeof_serial() function to insure that when we receive
a packet, we never end up calling the MiniportReturnPacket() routine
until after the receive handler has finished. I set things up so that
ndis_rxeof_serial() would only be used for serialized miniports since
they depend on this property. Well, it turns out deserialized miniports
depend on a similar property: you can't let MiniportReturnPacket() be
called from the same context as the receive handler at all. The 2100B
driver happens to use a single spinlock for all of its synchronization,
and it tries to acquire it both while in MiniportHandleInterrupt() and
in MiniportReturnPacket(), so if we call MiniportReturnPacket() from
the MiniportHandleInterrupt() context, we will end up trying to acquire
the spinlock recursively, which you can't do.
To fix this, I made the ndis_rxeof_serial() handler the default. An
alternate solution would be to make ndis_return_packet() submit
the call to MiniportReturnPacket() to the NDIS task queue thread.
I may do that in the future, after I've tested things a bit more.
attempting to duplicate Windows spinlocks. Windows spinlocks differ
from FreeBSD spinlocks in the way they block preemption. FreeBSD
spinlocks use critical_enter(), which masks off _all_ interrupts.
This prevents any other threads from being scheduled, but it also
prevents ISRs from running. In Windows, preemption is achieved by
raising the processor IRQL to DISPATCH_LEVEL, which prevents other
threads from preempting you, but does _not_ prevent device ISRs
from running. (This is essentially what Solaris calls dispatcher
locks.) The Windows spinlock itself (kspin_lock) is just an integer
value which is atomically set when you acquire the lock and atomically
cleared when you release it.
FreeBSD doesn't have IRQ levels, so we have to cheat a little by
using thread priorities: normal thread priority is PASSIVE_LEVEL,
lowest interrupt thread priority is DISPATCH_LEVEL, highest thread
priority is DEVICE_LEVEL (PI_REALTIME) and critical_enter() is
HIGH_LEVEL. In practice, only PASSIVE_LEVEL and DISPATCH_LEVEL
matter to us. The immediate benefit of all this is that I no
longer have to rely on a mutex pool.
Now, I'm sure many people will be seized by the urge to criticize
me for doing an end run around our own spinlock implementation, but
it makes more sense to do it this way. Well, it does to me anyway.
Overview of the changes:
- Properly implement hal_lock(), hal_unlock(), hal_irql(),
hal_raise_irql() and hal_lower_irql() so that they more closely
resemble their Windows counterparts. The IRQL is determined by
thread priority.
- Make ntoskrnl_lock_dpc() and ntoskrnl_unlock_dpc() do what they do
in Windows, which is to atomically set/clear the lock value. These
routines are designed to be called from DISPATCH_LEVEL, and are
actually half of the work involved in acquiring/releasing spinlocks.
- Add FASTCALL1(), FASTCALL2() and FASTCALL3() macros/wrappers
that allow us to call a _fastcall function in spite of the fact
that our version of gcc doesn't support __attribute__((__fastcall__))
yet. The macros take 1, 2 or 3 arguments, respectively. We need
to call hal_lock(), hal_unlock() etc... ourselves, but can't really
invoke the function directly. I could have just made the underlying
functions native routines and put _fastcall wrappers around them for
the benefit of Windows binaries, but that would create needless bloat.
- Remove ndis_mtxpool and all references to it. We don't need it
anymore.
- Re-implement the NdisSpinLock routines so that they use hal_lock()
and friends like they do in Windows.
- Use the new spinlock methods for handling lookaside lists and
linked list updates in place of the mutex locks that were there
before.
- Remove mutex locking from ndis_isr() and ndis_intrhand() since they're
already called with ndis_intrmtx held in if_ndis.c.
- Put ndis_destroy_lock() code under explicit #ifdef notdef/#endif.
It turns out there are some drivers which stupidly free the memory
in which their spinlocks reside before calling ndis_destroy_lock()
on them (touch-after-free bug). The ADMtek wireless driver
is guilty of this faux pas. (Why this doesn't clobber Windows I
have no idea.)
- Make NdisDprAcquireSpinLock() and NdisDprReleaseSpinLock() into
real functions instead of aliasing them to NdisAcaquireSpinLock()
and NdisReleaseSpinLock(). The Dpr routines use
KeAcquireSpinLockAtDpcLevel() level and KeReleaseSpinLockFromDpcLevel(),
which acquires the lock without twiddling the IRQL.
- In ndis_linksts_done(), do _not_ call ndis_80211_getstate(). Some
drivers may call the status/status done callbacks as the result of
setting an OID: ndis_80211_getstate() gets OIDs, which means we
might cause the driver to recursively access some of its internal
structures unexpectedly. The ndis_ticktask() routine will call
ndis_80211_getstate() for us eventually anyway.
- Fix the channel setting code a little in ndis_80211_setstate(),
and initialize the channel to IEEE80211_CHAN_ANYC. (The Microsoft
spec says you're not supposed to twiddle the channel in BSS mode;
I may need to enforce this later.) This fixes the problems I was
having with the ADMtek adm8211 driver: we were setting the channel
to a non-standard default, which would cause it to fail to associate
in BSS mode.
- Use hal_raise_irql() to raise our IRQL to DISPATCH_LEVEL when
calling certain miniport routines, per the Microsoft documentation.
I think that's everything. Hopefully, other than fixing the ADMtek
driver, there should be no apparent change in behavior.
case we should wait for the resetdone handler to be called before
returning.
- When providing resources via ndis_query_resources(), uses the
computed rsclen when using bcopy() to copy out the resource data
rather than the caller-supplied buffer length.
- Avoid using ndis_reset_nic() in if_ndis.c unless we really need
to reset the NIC because of a problem.
- Allow interrupts to be fielded during ndis_attach(), at least
as far as allowing ndis_isr() and ndis_intrhand() to run.
- Use ndis_80211_rates_ex when probing for supported rates. Technically,
this isn't supposed to work since, although Microsoft added the extended
rate structure with the NDIS 5.1 update, the spec still says that
the OID_802_11_SUPPORTED_RATES OID uses ndis_80211_rates. In spite of
this, it appears some drivers use it anyway.
- When adding in our guessed rates, check to see if they already exist
so that we avoid any duplicates.
- Add a printf() to ndis_open_file() that alerts the user when a
driver attempts to open a file under /compat/ndis.
With these changes, I can get the driver for the SMC 2802W 54g PCI
card to load and run. This board uses a Prism54G chip. Note that in
order for this driver to work, you must place the supplied smc2802w.arm
firmware image under /compat/ndis. (The firmware is not resident on
the device.)
Note that this should also allow the 3Com 3CRWE154G72 card to work
as well; as far as I can tell, these cards also use a Prism54G chip.
resource_var.h.
In kern_ndis.c:ndis_convert_res(), fill in the cprd_flags and
cprd_sharedisp fields as best we can.
In if_ndis.c:ndis_setmulti(), don't bother updating the multicast
filter if our multicast address list is empty.
Add some missing updates to ndis_var.h and ntoskrnl_var.h that I
forgot to check in when I added the KeDpc stuff.
variable length, so we should not be trying to copy it into a fixed
length buffer, especially one on the stack. malloc() a buffer of the
right size and return a pointer to that instead.
Fixes a crash I discovered when testing whe a Cisco AP in infrastructure
mode, which returns several information elements that make the
ndis_wlan_bssid_ex structure larger than expected.
(NIC would claim to establish a link with an ad-hoc net but it couldn't
send/receive packets). It turns out that every time the checkforhang
handler was called by ndis_ticktask(), the driver would generate a new
media connect event. The NDIS spec says the checkforhang handler is
called "approximately every 2 seconds" but using exactly 2 seconds seems
too fast. Using 3 seconds makes it happy again, so we'll go with that
for now.
are actually layered on top of the KeTimer API in subr_ntoskrnl.c, just
as it is in Windows. This reduces code duplication and more closely
imitates the way things are done in Windows.
- Modify ndis_encode_parm() to deal with the case where we have
a registry key expressed as a hex value ("0x1") which is being
read via NdisReadConfiguration() as an int. Previously, we tried
to decode things like "0x1" with strtol() using a base of 10, which
would always yield 0. This is what was causing problems with the
Intel 2200BG Centrino 802.11g driver: the .inf file that comes
with it has a key called RadioEnable with a value of 0x1. We
incorrectly decoded this value to '0' when it was queried, hence
the driver thought we wanted the radio turned off.
- In if_ndis.c, most drivers don't accept NDIS_80211_AUTHMODE_AUTO,
but NDIS_80211_AUTHMODE_SHARED may not be right in some cases,
so for now always use NDIS_80211_AUTHMODE_OPEN.
NOTE: There is still one problem with the Intel 2200BG driver: it
happens that the kernel stack in Windows is larger than the kernel
stack in FreeBSD. The 2200BG driver sometimes eats up more than 2
pages of stack space, which can lead to a double fault panic.
For the moment, I got things to work by adding the following to
my kernel config file:
options KSTACK_PAGES=8
I'm pretty sure 8 is too big; I just picked this value out of a hat
as a test, and it happened to work, so I left it. 4 pages might be
enough. Unfortunately, I don't think you can dynamically give a
thread a larger stack, so I'm not sure how to handle this short of
putting a note in the man page about it and dealing with the flood
of mail from people who never read man pages.
for Windows are deserialized miniports. Such drivers maintain their own
queues and do their own locking. This particular driver is not deserialized
though, and we need special support to handle it correctly.
Typically, in the ndis_rxeof() handler, we pass all incoming packets
directly to (*ifp->if_input)(). This in turn may cause another thread
to run and preempt us, and the packet may actually be processed and
then released before we even exit the ndis_rxeof() routine. The
problem with this is that releasing a packet calls the ndis_return_packet()
function, which hands the packet and its buffers back to the driver.
Calling ndis_return_packet() before ndis_rxeof() returns will screw
up the driver's internal queues since, not being deserialized,
it does no locking.
To avoid this problem, if we detect a serialized driver (by checking
the attribute flags passed to NdisSetAttributesEx(), we use an alternate
ndis_rxeof() handler, ndis_rxeof_serial(), which puts the call to
(*ifp->if_input)() on the NDIS SWI work queue. This guarantees the
packet won't be processed until after ndis_rxeof_serial() returns.
Note that another approach is to always copy the packet data into
another mbuf and just let the driver retain ownership of the ndis_packet
structure (ndis_return_packet() never needs to be called in this
case). I'm not sure which method is faster.
if_ndis.c has been split into if_ndis_pci.c and if_ndis_pccard.c.
The ndiscvt(8) utility should be able to parse device info for PCMCIA
devices now. The ndis_alloc_amem() has moved from kern_ndis.c to
if_ndis_pccard.c so that kern_ndis.c no longer depends on pccard.
NOTE: this stuff is not guaranteed to work 100% correctly yet. So
far I have been able to load/init my PCMCIA Cisco Aironet 340 card,
but it crashes in the interrupt handler. The existing support for
PCI/cardbus devices should still work as before.
that Asus provides on its CDs has both a MiniportSend() routine
and a MiniportSendPackets() function. The Microsoft NDIS docs say
that if a driver has both, only the MiniportSendPackets() routine
will be used. Although I think I implemented the support correctly,
calling the MiniportSend() routine seems to result in no packets going
out on the air, even though no error status is returned. The
MiniportSendPackets() function does work though, so at least in
this case it doesn't matter.
In if_ndis.c:ndis_getstate_80211(), if ndis_get_assoc() returns
an error, don't bother trying to obtain any other state since the
calls may fail, or worse cause the underlying driver to crash.
(The above two changes make the Asus-supplied Centrino work.)
Also, when calling the OID_802_11_CONFIGURATION OID, remember
to initialize the structure lengths correctly.
In subr_ndis.c:ndis_open_file(), set the current working directory
to rootvnode if we're in a thread that doesn't have a current
working directory set.
is for an 802.11 device or not. At least one driver I have does not
support the OID_802_11_NETWORK_TYPES_SUPPORTED OID.
Also, for now, don't do anything special in the ndis_suspend() method.
I originally wanted to shut down the NIC but leave the IFF_UP flag alone
since technically the interface is meant to remain up, but an interrupt
may be delivered to the ISR on suspend, and if this happens while the
NIC is halted, we will crash, since none of the miniport driver methods
will function.
This needs to be dealt with properly later, but for now this prevents
a panic, and the resume method properly re-inits the NIC.
won't associate in BSS mode if you use AUTHMODE_SHARED. I probably don't
understand enough to know when SHARED should be used vs. OPEN or WPA.
For now, go back to what works.
on an SIOCSIFADDR (by way of brain damage in net80211).
Also, avoid trying to set NDIS_80211_AUTHMODE_AUTO since the Microsoft
documentation I have recommends not using it, and the Centrino driver
seems to dislike being told to use it.
