spinlock & event question

NTDEV
1

Hi all,

Is it possible to both release a spinlock and wait for a notification
event, atomically, as a single operation?
Does the kernel offer such support, and if yes, with which routine?

thank you very much,

Sandor LUKACS
Virus Analyst, SOFTWIN

ps. it might be a very simple question, but I’m looking for
such a function for over a hour and just doesn’t see it :frowning:

2

Is not possible to wait for anything as a atomic operation.

----- Original Message -----
From:
To: “Windows System Software Devs Interest List”
Sent: Monday, November 06, 2006 10:47 AM
Subject: [ntdev] spinlock & event question

> Hi all,
>
> Is it possible to both release a spinlock and wait for a notification
> event, atomically, as a single operation?
> Does the kernel offer such support, and if yes, with which routine?
>
> thank you very much,
>
> Sandor LUKACS
> Virus Analyst, SOFTWIN
>
> ps. it might be a very simple question, but I’m looking for
> such a function for over a hour and just doesn’t see it :frowning:
>
> —
> Questions? First check the Kernel Driver FAQ at
> http://www.osronline.com/article.cfm?id=256
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer

3

Hi,

Is not possible to wait for anything as a atomic operation.

AFAIK it IS possible to wait atomically for semaphores; take a look at
KeReleaseSemaphore, if you call it with Wait = TRUE, then you can
release the semaphore and call immediatly a KeWaitFor… function
in atomical fashion;

but I would need to do such a thing, but with releasing a spinlock
instead of a semaphore; any other ideas?

thank you,

Sandor LUKACS
Virus Analyst, SOFTWIN

4

Don;t think you can find a solution for this.

~Sisimon

On 11/6/06, xxxxx@bitdefender.com wrote:
>
> Hi all,
>
> Is it possible to both release a spinlock and wait for a notification
> event, atomically, as a single operation?
> Does the kernel offer such support, and if yes, with which routine?
>
> thank you very much,
>
> Sandor LUKACS
> Virus Analyst, SOFTWIN
>
> ps. it might be a very simple question, but I’m looking for
> such a function for over a hour and just doesn’t see it :frowning:
>
> —
> Questions? First check the Kernel Driver FAQ at
> http://www.osronline.com/article.cfm?id=256
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
>

–
GCS d+ s: a- c++++ U> B+ L++>$ w++++$ W++(+++) PGP+N+ t PS+PE++ tv+(++) b+++
G+++ e++>(++++) h-- r
Don’t know this? See http://www.geekcode.com/geek.html

5

I understand that you are trying to build the same posix semantics for
your spinlocks.
You can try the following (AT YOUR RISK!!) with a caveat that you cannot
use MYSPINLOCK_SLEEP at dispatch level.

Declare your spinlock struct as
typedef struct {
KSPIN_LOCK SpinLock; /* spin lock */
KSEMAPHORE Sem; /* semaphore object */
ULONG nsleepers; /* # of threads sleeping
*/
ULONG nwakeups; /* # of wakeups */
} my_spinlock_t;

MYSPINLOCK_LOCK simply aquires the SpinLock and MYSPINLOCK_UNLOCK
releases the SpinLock

MYSPINLOCK_SLEEP(lock) macro /* assumes that lock is already aquired */
do {
KIRQL zopri;
++(lock)->nsleepers; \
zopri = PASSIVE_LEVEL; \
MYSPINLOCK_UNLOCK(lock, zopri); \
KeWaitForSingleObject(&(lock)->Sem, \
Executive, KernelMode, FALSE, NULL);\
MYSPINLOCK_LOCK(lock, zopri); \
–(lock)->nsleepers; \
–(lock)->nwakeups; \
} while (0)

MYSPINLOCK_WAKEUP(lock) macro to wakeup all those sleeping
do {
if ((lock)->nsleepers > (lock)->nwakeups) { \
KeReleaseSemaphore(&(lock)->Sem, \
IO_NO_INCREMENT, \
(lock)->nsleepers - \
(lock)->nwakeups, FALSE); \
(lock)->nwakeups = (lock)->nsleepers; \
} \
} while (0)

MYSPINLOCK_WAKEUP_ONE would be a variant in which you would release only
1 Semaphore resource.

-----Original Message-----
From: xxxxx@lists.osr.com [mailto:bounce-268957-
xxxxx@lists.osr.com] On Behalf Of xxxxx@bitdefender.com
Sent: Monday, November 06, 2006 12:47 AM
To: Windows System Software Devs Interest List
Subject: [ntdev] spinlock & event question

Hi all,

Is it possible to both release a spinlock and wait for a notification
event, atomically, as a single operation?
Does the kernel offer such support, and if yes, with which routine?

thank you very much,

Sandor LUKACS
Virus Analyst, SOFTWIN

ps. it might be a very simple question, but I’m looking for
such a function for over a hour and just doesn’t see it :frowning:

Questions? First check the Kernel Driver FAQ at
http://www.osronline.com/article.cfm?id=256

To unsubscribe, visit the List Server section of OSR Online at
http://www.osronline.com/page.cfm?name=ListServer

6

On Nov 6, 2006, at 12:46 PM, Harish Arora wrote:

I understand that you are trying to build the same posix semantics for
your spinlocks.
You can try the following (AT YOUR RISK!!) with a caveat that you
cannot
use MYSPINLOCK_SLEEP at dispatch level.

Declare your spinlock struct as
typedef struct {
KSPIN_LOCK SpinLock; /* spin lock */
KSEMAPHORE Sem; /* semaphore object */
ULONG nsleepers; /* # of threads sleeping
*/
ULONG nwakeups; /* # of wakeups */
} my_spinlock_t;

MYSPINLOCK_LOCK simply aquires the SpinLock and MYSPINLOCK_UNLOCK
releases the SpinLock

MYSPINLOCK_SLEEP(lock) macro /* assumes that lock is already
aquired */
do {
KIRQL zopri;
++(lock)->nsleepers; \
zopri = PASSIVE_LEVEL; \
MYSPINLOCK_UNLOCK(lock, zopri); \
KeWaitForSingleObject(&(lock)->Sem, \
Executive, KernelMode, FALSE, NULL);\
MYSPINLOCK_LOCK(lock, zopri); \
–(lock)->nsleepers; \
–(lock)->nwakeups; \
} while (0)

This looks broken to me; you release the spin lock before you wait;
that’s not atomic at all. You’re going to need OS support for this.

-sd

7

Hi all,

Can anyone tell me if and how could I check from a driver, that a process which tries to connect to a communication port of the driver is digitally signed or not, if the signature if valid, who signed that process and so on. This could be a quite effective way to prevent unauthorized processes to connect to the driver.

Any comments on this subject are welcomed.

thank you very much,

Sandor LUKACS
Virus Analyst, SOFTWIN

8

sorry about the prior post, should have been a NEW topic,

Sandor LUKACS

9

How do you know that someone hasn’t injected code into your process with a
signed .exe?

This approach is flawed.

–
Ken Johnson (Skywing)
Windows SDK MVP
http://www.nynaeve.net

wrote in message news:xxxxx@ntdev…
> Hi all,
>
> Can anyone tell me if and how could I check from a driver, that a process
> which tries to connect to a communication port of the driver is digitally
> signed or not, if the signature if valid, who signed that process and so
> on. This could be a quite effective way to prevent unauthorized processes
> to connect to the driver.
>
> Any comments on this subject are welcomed.
>
> thank you very much,
>
> Sandor LUKACS
> Virus Analyst, SOFTWIN
>