Skip to contentSkip to conversation

mct_u232: added _ioctl, _msr_to_icount and _get_icount functions

PATCH Tsozik 3 participants 7 messages Dec 27, 2010 – Jan 5, 2011

lore
Share & tools
lore

Tsozik

author
#0 lore raw

Pete,

You're absolutely right. My apology again, I re-checked values of MSR masks used to update icount counters and they all belong to upper MSR nibble, so they indeed track changes (not states). I took out counter increments from mct_u232_msr_to_state function and encapsulated them in the newly defined mct_u232_msr_to_icount function, effectively leaving the previous implementation of mct_u232_msr_to_state function intact. mct_u232_msr_to_icount uses delta nibble to track state changes. RM-60 testing showed the same levels as measured by PDM-2 in close proximity to RM-60. Please let me know if anything else needs to be corrected to roll this out,

Thank you again for your expertise,
Vadim.

Quoted 115 lines · Pete Zaitcev

From: Pete Zaitcev <[email protected]>
Subject: Re: [PATCH 1/1] mct_u232: added _ioctl and _get_icount functions
To: "Tsozik" <[email protected]>
Cc: "Greg Kroah-Hartman" <[email protected]>, [email protected], [email protected]
Date: Monday, December 27, 2010, 12:08 PM
On Sun, 26 Dec 2010 22:34:23 -0800
(PST)
Tsozik <[email protected]>

Quoted 1 line · wrote:

Usually MSR contains information about both states

which are raised and

Quoted 1 line

states which are changed. It looks like we use upper

MSR nibble which

Quoted 1 line

contains information about 4 states which were raised,

not lower MSR

Quoted 1 line

nibble which contains information about 4 states which

were changed.

Sure, I know how typical UART works. Does the MCT device deliver the
change states? If it does and you want to use those, go ahead.

Quoted 1 line

I also researched implementations under usb/serial and

saw that all of

Quoted 1 line

them are rely on the current states (not change

states) to increment

Quoted 1 line

the respective counter variables.

Here's drivers/usb/serial/io_edgeport.c:

static void handle_new_msr(struct edgeport_port *edge_port, __u8 newMsr)
{
    struct  async_icount *icount;

    dbg("%s %02x", __func__, newMsr);

    if (newMsr & (EDGEPORT_MSR_DELTA_CTS
| EDGEPORT_MSR_DELTA_DSR |

EDGEPORT_MSR_DELTA_RI | EDGEPORT_MSR_DELTA_CD)) {
        icount =
&edge_port->icount;

        /* update input line
counters */
        if (newMsr & EDGEPORT_MSR_DELTA_CTS)

icount->cts++;
        if (newMsr & EDGEPORT_MSR_DELTA_DSR)

icount->dsr++;
        if (newMsr & EDGEPORT_MSR_DELTA_CD)

icount->dcd++;
        if (newMsr & EDGEPORT_MSR_DELTA_RI)

icount->rng++;
wake_up_interruptible(&edge_port->delta_msr_wait); }

Looks like it counts changes to me. And in any case, the gold standard
is in drivers/serial/8250.c:

static unsigned int check_modem_status(struct
uart_8250_port *up)
{
    unsigned int status = serial_in(up,
UART_MSR);

    status |= up->msr_saved_flags;
    up->msr_saved_flags = 0;
    if (status & UART_MSR_ANY_DELTA
&& up->ier & UART_IER_MSI &&
        up->port.state != NULL)
{
        if (status &
UART_MSR_TERI)

up->port.icount.rng++;
        if (status &
UART_MSR_DDSR)

up->port.icount.dsr++;
        if (status &
UART_MSR_DDCD)

uart_handle_dcd_change(&up->port, status &
UART_MSR_DCD);
        if (status &
UART_MSR_DCTS)

uart_handle_cts_change(&up->port, status &
UART_MSR_CTS);

wake_up_interruptible(&up->port.state->port.delta_msr_wait); }

I only proposed accomplishing the same result by comparing states
in place of relying on hardware change reports like both of the
above do.

-- Pete

#1 lore raw

Review

Quoted 13 lines · On Mon, 27 Dec 2010 12:55:38 -0800 (PST) Tsozik <[email protected]> wrote:
(quoted diff) +13 -0
+static void mct_u232_msr_to_icount(struct async_icount *icount,
+						unsigned char msr)
+{
+	/* Translate Control Line states */
+	if (msr & MCT_U232_MSR_DDSR)
+		icount->dsr++;
+	if (msr & MCT_U232_MSR_DCTS)
+		icount->cts++;
+	if (msr & MCT_U232_MSR_DRI)
+		icount->rng++;
+	if (msr & MCT_U232_MSR_DCD)
+		icount->dcd++;
+} /* mct_u232_msr_to_icount */

This looks good to me, but since it's a new hardware facility that we did not use previously, I'd like to make sure this works for you.

Quoted 3 lines

mct_u232_msr_to_icount uses delta nibble to track state changes. RM-60 testing showed the same levels as measured by PDM-2 in close proximity to RM-60.

Right, this is good. I have just one request for you: could you find somewhere a computer with a built-in 8250-compatible serial port, attach RM-60 to it, and run your counting application there? I think your patch looks just fine, but I would like to make sure that we (e.g. mct_u232+patch) are truly compatible now.

-- Pete

author
#2 lore raw

Pete,

Many thanks for your reply again. I actually still own an antique HP a230n (disclaimer: I got it from/with my wife and would never buy it myself). According to dmesg it's got built-in 8250-compliant serial:

[vtsozik@f15 serial]$ dmesg | grep -i serial | grep 8250
Serial: 8250/16550 driver, 4 ports, IRQ sharing enabled
serial8250: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A [vtsozik@f15 serial]$

So I ran geiger counter against /dev/ttyS0 device for 20 minutes and acquired 20 measurements. Then I compared last average with last 20 minute measurement average acquired via mct_u232 on the laptop placed nearby. The error was ~4% (rounded up). Taking in consideration that measurements were not acquired synchronously (i.e. in 2 different time slots) this is a pretty good match.

Thank you again,
Vadim.

Quoted 55 lines · Pete Zaitcev

From: Pete Zaitcev <[email protected]>
Subject: Re: [PATCH 1/1] mct_u232: added _ioctl, _msr_to_icount and _get_icount functions
To: "Tsozik" <[email protected]>
Cc: "Greg Kroah-Hartman" <[email protected]>, [email protected], [email protected], [email protected]
Date: Monday, December 27, 2010, 5:12 PM
On Mon, 27 Dec 2010 12:55:38 -0800
(PST)
Tsozik <[email protected]>

Quoted 1 line · wrote:

+static void mct_u232_msr_to_icount(struct

async_icount *icount,

Quoted 1 line

+


    unsigned char msr)

Quoted 2 lines

+{
+    /* Translate Control Line states

*/

Quoted 2 lines

+    if (msr & MCT_U232_MSR_DDSR)
+

icount->dsr++;

Quoted 2 lines

+    if (msr & MCT_U232_MSR_DCTS)
+

icount->cts++;

Quoted 2 lines

+    if (msr & MCT_U232_MSR_DRI)
+

icount->rng++;

Quoted 2 lines

+    if (msr & MCT_U232_MSR_DCD)
+

icount->dcd++;

Quoted 1 line

+} /* mct_u232_msr_to_icount */

This looks good to me, but since it's a new hardware facility that
we did not use previously, I'd like to make sure this works for you.

Quoted 1 line

mct_u232_msr_to_icount uses delta nibble to track

state changes.

Quoted 1 line

RM-60 testing showed the same levels as measured by

PDM-2 in close

Quoted 1 line

proximity to RM-60.

Right, this is good. I have just one request for you: could you find
somewhere a computer with a built-in 8250-compatible serial port,
attach RM-60 to it, and run your counting application there?
I think your patch looks just fine, but I would like to make sure
that we (e.g. mct_u232+patch) are truly compatible now.

-- Pete

#3 lore raw
Quoted 4 lines · On Mon, 27 Dec 2010 20:04:51 -0800 (PST) Tsozik <[email protected]> wrote:

So I ran geiger counter against /dev/ttyS0 device for 20 minutes and acquired 20 measurements. Then I compared last average with last 20 minute measurement average acquired via mct_u232 on the laptop placed nearby. The error was ~4% (rounded up).

Great, I'm ready to ack.

There's just one thing that is bugging me... I think it would be best if Alan Cox or Greg Kroah commented on it. The edgeport does the following, which we copied:

		schedule();
		........
		if (cnow.rng == cprev.rng && cnow.dsr == cprev.dsr &&
		    cnow.dcd == cprev.dcd && cnow.cts == cprev.cts)
			return -EIO; /* no change => error */
		if (((arg & TIOCM_RNG) && (cnow.rng != cprev.rng)) ||
		    ((arg & TIOCM_DSR) && (cnow.dsr != cprev.dsr)) ||
		    ((arg & TIOCM_CD)  && (cnow.dcd != cprev.dcd)) ||
		    ((arg & TIOCM_CTS) && (cnow.cts != cprev.cts))) {
			return 0;
		}

So, if there was a status report, but no change to bits, the ioctl TIOCMIWAIT would return with -EIO. In serial_core.c, that serves conventional non-USB UARTs, nothing like this occurs. I am not quite sure what the point of doing this -EIO check is.

Oh and BTW, I'm wondering what is going to happen if the device is disconnected while an application is blocked waiting for the status change. The patch is not particularly bad here, it just copies an existing code from elsewhere.

-- Pete

author
#4 lore raw

Greg,

I'm sorry to bother you again, but I'm wondering if you could comment on Pete's concern below.

Thank you in advance for your expertise on the matter,
Vadim.

Quoted 66 lines · Pete Zaitcev

From: Pete Zaitcev <[email protected]>
Subject: Re: [PATCH 1/1] mct_u232: added _ioctl, _msr_to_icount and _get_icount functions
To: "Tsozik" <[email protected]>
Cc: "Greg Kroah-Hartman" <[email protected]>, [email protected], [email protected], [email protected]
Date: Tuesday, December 28, 2010, 1:40 AM
On Mon, 27 Dec 2010 20:04:51 -0800
(PST)
Tsozik <[email protected]>

Quoted 1 line · wrote:

So I ran geiger counter against /dev/ttyS0 device for

20 minutes and

Quoted 1 line

acquired 20 measurements. Then I compared last average

with last 20

Quoted 1 line

minute measurement average acquired via mct_u232 on

the laptop placed

Quoted 1 line

nearby. The error was ~4% (rounded up).

Great, I'm ready to ack.

There's just one thing that is bugging me... I think it would be best
if Alan Cox or Greg Kroah commented on it. The edgeport does the
following, which we copied:

        schedule();
        ........
        if (cnow.rng ==
cprev.rng && cnow.dsr == cprev.dsr &&

cnow.dcd == cprev.dcd && cnow.cts == cprev.cts)

return -EIO; /* no change => error */
        if (((arg &
TIOCM_RNG) && (cnow.rng != cprev.rng)) ||
            ((arg
& TIOCM_DSR) && (cnow.dsr != cprev.dsr)) ||
            ((arg
& TIOCM_CD)  && (cnow.dcd != cprev.dcd))
||
            ((arg
& TIOCM_CTS) && (cnow.cts != cprev.cts))) {

return 0;
        }

So, if there was a status report, but no change to bits, the ioctl
TIOCMIWAIT would return with -EIO. In serial_core.c, that serves
conventional non-USB UARTs, nothing like this occurs. I am not quite
sure what the point of doing this -EIO check is.

Oh and BTW, I'm wondering what is going to happen if the device is
disconnected while an application is blocked waiting for the status
change. The patch is not particularly bad here, it just copies
an existing code from elsewhere.

-- Pete

#5 lore raw
In reply to Pete Zaitcev
Quoted 27 lines · Pete Zaitcev
Quoted 4 lines · On Mon, 27 Dec 2010 20:04:51 -0800 (PST) Tsozik <[email protected]> wrote:

So I ran geiger counter against /dev/ttyS0 device for 20 minutes and acquired 20 measurements. Then I compared last average with last 20 minute measurement average acquired via mct_u232 on the laptop placed nearby. The error was ~4% (rounded up).

Great, I'm ready to ack.

There's just one thing that is bugging me... I think it would be best if Alan Cox or Greg Kroah commented on it. The edgeport does the following, which we copied:

		schedule();
		........
		if (cnow.rng == cprev.rng && cnow.dsr == cprev.dsr &&
		    cnow.dcd == cprev.dcd && cnow.cts == cprev.cts)
			return -EIO; /* no change => error */
		if (((arg & TIOCM_RNG) && (cnow.rng != cprev.rng)) ||
		    ((arg & TIOCM_DSR) && (cnow.dsr != cprev.dsr)) ||
		    ((arg & TIOCM_CD)  && (cnow.dcd != cprev.dcd)) ||
		    ((arg & TIOCM_CTS) && (cnow.cts != cprev.cts))) {
			return 0;
		}
Quoted 4 lines

So, if there was a status report, but no change to bits, the ioctl TIOCMIWAIT would return with -EIO. In serial_core.c, that serves conventional non-USB UARTs, nothing like this occurs. I am not quite sure what the point of doing this -EIO check is.

Odd, I don't really remember where I copied that logic from, that was a long time ago.

Quoted 4 lines

Oh and BTW, I'm wondering what is going to happen if the device is disconnected while an application is blocked waiting for the status change. The patch is not particularly bad here, it just copies an existing code from elsewhere.

It should handle the disconnect, if not, we should fix it.

thanks,

greg k-h

#6 lore raw
In reply to Tsozik
Quoted 3 lines · Tsozik

Pete,

You're absolutely right. My apology again, I re-checked values of MSR masks used to update icount counters and they all belong to upper MSR nibble, so they indeed track changes (not states). I took out counter increments from mct_u232_msr_to_state function and encapsulated them in the newly defined mct_u232_msr_to_icount function, effectively leaving the previous implementation of mct_u232_msr_to_state function intact. mct_u232_msr_to_icount uses delta nibble to track state changes. RM-60 testing showed the same levels as measured by PDM-2 in close proximity to RM-60. Please let me know if anything else needs to be corrected to roll this out,

Can you not attach the patch as a base64 attachment? I can't apply it as-is, because of that, sorry.

thanks,

greg k-h