From: "V. Ananda Krishnan" <mansarov@us.ibm.com>
To: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: linux-kernel@vger.kernel.org, rmk@arm.linux.org.uk,
akpm@osdl.org, gregkh@suse.de
Subject: Re: [PATCH]-jsm driver fix for linux-2.6.16-rc1
Date: Mon, 23 Jan 2006 12:27:47 -0600 [thread overview]
Message-ID: <43D52023.4090607@us.ibm.com> (raw)
In-Reply-To: <1137781399.24161.30.camel@localhost.localdomain>
[-- Attachment #1: Type: text/plain, Size: 789 bytes --]
I am submitting the revised patch. We removed the use of jsm-rawreadok
feature. Greg's suggestions are also taken care of. Let me have the
feed-back.
Thanks,
V. Ananda Krishnan
Alan Cox wrote:
> On Gwe, 2006-01-20 at 10:02 -0600, V. Ananda Krishnan wrote:
>> Hi,
>>
>> The following patch takes into account the dynamically allocated
>> tty_buf changes and hence fixes the jsm driver. Please let me have the
>> feed-back.
>
> - Do you still need the shortcut stuff with your own private buffer ?
>
> I'd like to move to a model where ldiscs free up the tty buffers to both
> improve the locking and kill the horrible DONT_FLIP hacks in n_tty. If
> you do it isn't a problem but I need to ensure there is a way to
> directly allocate a buffer for this use.
>
> Alan
>
>
[-- Attachment #2: patch_jsm_012306.txt --]
[-- Type: text/plain, Size: 8736 bytes --]
diff -Naur linux-2.6.16-rc1/drivers/serial/Kconfig linux-2.6.16-rc1-mod/drivers/serial/Kconfig
--- linux-2.6.16-rc1/drivers/serial/Kconfig 2006-01-23 11:08:19.000000000 -0600
+++ linux-2.6.16-rc1-mod/drivers/serial/Kconfig 2006-01-23 11:06:41.000000000 -0600
@@ -893,7 +893,7 @@
config SERIAL_JSM
tristate "Digi International NEO PCI Support"
- depends on PCI && BROKEN
+ depends on PCI
select SERIAL_CORE
help
This is a driver for Digi International's Neo series
diff -Naur linux-2.6.16-rc1/drivers/serial/jsm/jsm_tty.c linux-2.6.16-rc1-mod/drivers/serial/jsm/jsm_tty.c
--- linux-2.6.16-rc1/drivers/serial/jsm/jsm_tty.c 2006-01-23 11:03:24.000000000 -0600
+++ linux-2.6.16-rc1-mod/drivers/serial/jsm/jsm_tty.c 2006-01-23 11:25:42.000000000 -0600
@@ -20,8 +20,10 @@
*
* Contact Information:
* Scott H Kilau <Scott_Kilau@digi.com>
- * Wendy Xiong <wendyx@us.ltcfwd.linux.ibm.com>
- *
+ * Ananda Venkatarman <mansarov@us.ibm.com>
+ * Modifications:
+ * 01/19/06: changed jsm_input routine to use the dynamically allocated
+ * tty_buffer changes. Contributors: Scott Kilau and Ananda V.
***********************************************************************/
#include <linux/tty.h>
#include <linux/tty_flip.h>
@@ -497,16 +499,16 @@
{
struct jsm_board *bd;
struct tty_struct *tp;
+ struct tty_ldisc *ld;
u32 rmask;
u16 head;
u16 tail;
int data_len;
unsigned long lock_flags;
- int flip_len;
+ int flip_len = 0;
int len = 0;
int n = 0;
char *buf = NULL;
- char *buf2 = NULL;
int s = 0;
int i = 0;
@@ -574,59 +576,54 @@
/*
* If the rxbuf is empty and we are not throttled, put as much
- * as we can directly into the linux TTY flip buffer.
- * The jsm_rawreadok case takes advantage of carnal knowledge that
- * the char_buf and the flag_buf are next to each other and
- * are each of (2 * TTY_FLIPBUF_SIZE) size.
+ * as we can directly into the linux TTY buffer.
*
- * NOTE: if(!tty->real_raw), the call to ldisc.receive_buf
- *actually still uses the flag buffer, so you can't
- *use it for input data
- */
- if (jsm_rawreadok) {
- if (tp->real_raw)
- flip_len = MYFLIPLEN;
- else
- flip_len = 2 * TTY_FLIPBUF_SIZE;
- } else
- flip_len = TTY_FLIPBUF_SIZE - tp->flip.count;
+ */
+ flip_len = TTY_FLIPBUF_SIZE;
len = min(data_len, flip_len);
len = min(len, (N_TTY_BUF_SIZE - 1) - tp->read_cnt);
+ ld = tty_ldisc_ref(tp);
+
+ /*
+ * If the DONT_FLIP flag is on, don't flush our buffer, and act
+ * like the ld doesn't have any space to put the data right now.
+ */
+ if (test_bit(TTY_DONT_FLIP, &tp->flags))
+ len = 0;
+
+ /*
+ * If we were unable to get a reference to the ld,
+ * don't flush our buffer, and act like the ld doesn't
+ * have any space to put the data right now.
+ */
+ if (!ld) {
+ len = 0;
+ } else {
+ /*
+ * If ld doesn't have a pointer to a receive_buf function,
+ * flush the data, then act like the ld doesn't have any
+ * space to put the data right now.
+ */
+ if (!ld->receive_buf) {
+ ch->ch_r_head = ch->ch_r_tail;
+ len = 0;
+ }
+ }
+
if (len <= 0) {
spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
jsm_printk(READ, INFO, &ch->ch_bd->pci_dev, "jsm_input 1\n");
+ if (ld)
+ tty_ldisc_deref(ld);
return;
}
- /*
- * If we're bypassing flip buffers on rx, we can blast it
- * right into the beginning of the buffer.
- */
- if (jsm_rawreadok) {
- if (tp->real_raw) {
- if (ch->ch_flags & CH_FLIPBUF_IN_USE) {
- jsm_printk(READ, INFO, &ch->ch_bd->pci_dev,
- "JSM - FLIPBUF in use. delaying input\n");
- spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
- return;
- }
- ch->ch_flags |= CH_FLIPBUF_IN_USE;
- buf = ch->ch_bd->flipbuf;
- buf2 = NULL;
- } else {
- buf = tp->flip.char_buf;
- buf2 = tp->flip.flag_buf;
- }
- } else {
- buf = tp->flip.char_buf_ptr;
- buf2 = tp->flip.flag_buf_ptr;
- }
-
+ len = tty_buffer_request_room(tp, len);
n = len;
- /*
+ /*
* n now contains the most amount of data we can copy,
* bounded either by the flip buffer size or the amount
* of data the card actually has pending...
@@ -638,121 +635,47 @@
if (s <= 0)
break;
- memcpy(buf, ch->ch_rqueue + tail, s);
-
- /* buf2 is only set when port isn't raw */
- if (buf2)
- memcpy(buf2, ch->ch_equeue + tail, s);
-
- tail += s;
- buf += s;
- if (buf2)
- buf2 += s;
- n -= s;
- /* Flip queue if needed */
- tail &= rmask;
- }
+ /*
+ * If conditions are such that ld needs to see all
+ * UART errors, we will have to walk each character
+ * and error byte and send them to the buffer one at
+ * a time.
+ */
- /*
- * In high performance mode, we don't have to update
- * flag_buf or any of the counts or pointers into flip buf.
- */
- if (!jsm_rawreadok) {
if (I_PARMRK(tp) || I_BRKINT(tp) || I_INPCK(tp)) {
- for (i = 0; i < len; i++) {
+ for (i = 0; i < s; i++) {
/*
* Give the Linux ld the flags in the
* format it likes.
*/
- if (tp->flip.flag_buf_ptr[i] & UART_LSR_BI)
- tp->flip.flag_buf_ptr[i] = TTY_BREAK;
- else if (tp->flip.flag_buf_ptr[i] & UART_LSR_PE)
- tp->flip.flag_buf_ptr[i] = TTY_PARITY;
- else if (tp->flip.flag_buf_ptr[i] & UART_LSR_FE)
- tp->flip.flag_buf_ptr[i] = TTY_FRAME;
+ if (*(ch->ch_equeue +tail +i) & UART_LSR_BI)
+ tty_insert_flip_char(tp, *(ch->ch_rqueue +tail +i), TTY_BREAK);
+ else if (*(ch->ch_equeue +tail +i) & UART_LSR_PE)
+ tty_insert_flip_char(tp, *(ch->ch_rqueue +tail +i), TTY_PARITY);
+ else if (*(ch->ch_equeue +tail +i) & UART_LSR_FE)
+ tty_insert_flip_char(tp, *(ch->ch_rqueue +tail +i), TTY_FRAME);
else
- tp->flip.flag_buf_ptr[i] = TTY_NORMAL;
- }
+ tty_insert_flip_char(tp, *(ch->ch_rqueue +tail +i), TTY_NORMAL);
+ }
} else {
- memset(tp->flip.flag_buf_ptr, 0, len);
+ tty_insert_flip_string(tp, ch->ch_rqueue + tail, s) ;
}
-
- tp->flip.char_buf_ptr += len;
- tp->flip.flag_buf_ptr += len;
- tp->flip.count += len;
- }
- else if (!tp->real_raw) {
- if (I_PARMRK(tp) || I_BRKINT(tp) || I_INPCK(tp)) {
- for (i = 0; i < len; i++) {
- /*
- * Give the Linux ld the flags in the
- * format it likes.
- */
- if (tp->flip.flag_buf_ptr[i] & UART_LSR_BI)
- tp->flip.flag_buf_ptr[i] = TTY_BREAK;
- else if (tp->flip.flag_buf_ptr[i] & UART_LSR_PE)
- tp->flip.flag_buf_ptr[i] = TTY_PARITY;
- else if (tp->flip.flag_buf_ptr[i] & UART_LSR_FE)
- tp->flip.flag_buf_ptr[i] = TTY_FRAME;
- else
- tp->flip.flag_buf_ptr[i] = TTY_NORMAL;
- }
- } else
- memset(tp->flip.flag_buf, 0, len);
+ tail += s;
+ n -= s;
+ /* Flip queue if needed */
+ tail &= rmask;
}
- /*
- * If we're doing raw reads, jam it right into the
- * line disc bypassing the flip buffers.
- */
- if (jsm_rawreadok) {
- if (tp->real_raw) {
- ch->ch_r_tail = tail & rmask;
- ch->ch_e_tail = tail & rmask;
-
- jsm_check_queue_flow_control(ch);
-
- /* !!! WE *MUST* LET GO OF ALL LOCKS BEFORE CALLING RECEIVE BUF !!! */
+ ch->ch_r_tail = tail & rmask;
+ ch->ch_e_tail = tail & rmask;
+ jsm_check_queue_flow_control(ch);
+ spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
- spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
+ /* Tell the tty layer its okay to "eat" the data now */
+ tty_flip_buffer_push(tp);
- jsm_printk(READ, INFO, &ch->ch_bd->pci_dev,
- "jsm_input. %d real_raw len:%d calling receive_buf for board %d\n",
- __LINE__, len, ch->ch_bd->boardnum);
- tp->ldisc.receive_buf(tp, ch->ch_bd->flipbuf, NULL, len);
-
- /* Allow use of channel flip buffer again */
- spin_lock_irqsave(&ch->ch_lock, lock_flags);
- ch->ch_flags &= ~CH_FLIPBUF_IN_USE;
- spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
-
- } else {
- ch->ch_r_tail = tail & rmask;
- ch->ch_e_tail = tail & rmask;
-
- jsm_check_queue_flow_control(ch);
-
- /* !!! WE *MUST* LET GO OF ALL LOCKS BEFORE CALLING RECEIVE BUF !!! */
- spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
-
- jsm_printk(READ, INFO, &ch->ch_bd->pci_dev,
- "jsm_input. %d not real_raw len:%d calling receive_buf for board %d\n",
- __LINE__, len, ch->ch_bd->boardnum);
-
- tp->ldisc.receive_buf(tp, tp->flip.char_buf, tp->flip.flag_buf, len);
- }
- } else {
- ch->ch_r_tail = tail & rmask;
- ch->ch_e_tail = tail & rmask;
-
- jsm_check_queue_flow_control(ch);
-
- spin_unlock_irqrestore(&ch->ch_lock, lock_flags);
-
- jsm_printk(READ, INFO, &ch->ch_bd->pci_dev,
- "jsm_input. %d not jsm_read raw okay scheduling flip\n", __LINE__);
- tty_schedule_flip(tp);
- }
+ if (ld)
+ tty_ldisc_deref(ld);
jsm_printk(IOCTL, INFO, &ch->ch_bd->pci_dev, "finish\n");
}
next prev parent reply other threads:[~2006-01-23 18:26 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-01-20 16:02 [PATCH]-jsm driver fix for linux-2.6.16-rc1 V. Ananda Krishnan
2006-01-20 16:10 ` Greg KH
2006-01-20 18:23 ` Alan Cox
2006-01-23 18:27 ` V. Ananda Krishnan [this message]
2006-01-23 18:46 ` Greg KH
2006-01-23 21:19 ` V. Ananda Krishnan
2006-01-23 21:42 ` Greg KH
2006-01-24 22:28 ` Greg KH
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=43D52023.4090607@us.ibm.com \
--to=mansarov@us.ibm.com \
--cc=akpm@osdl.org \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=gregkh@suse.de \
--cc=linux-kernel@vger.kernel.org \
--cc=rmk@arm.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox