diff mbox series

[2/2] usb: gadget: ncm: Add support to update wMaxSegmentSize via configfs

Message ID 20231009142005.21338-2-quic_kriskura@quicinc.com
State New
Headers show
Series None | expand

Commit Message

Krishna Kurapati Oct. 9, 2023, 2:20 p.m. UTC
Currently the NCM driver restricts wMaxSegmentSize that indicates
the datagram size coming from network layer to 1514. However the
spec doesn't have any limitation. For P2P connections over NCM,
increasing MTU helps increasing throughput.

Add support to configure this value before configfs symlink is
created. Also since the NTB Out/In buffer sizes are fixed at 16384
bytes, limit the segment size to an upper cap of 15014. Set the
default MTU size for the ncm interface during function bind before
network interface is registered allowing MTU to be set in parity
with wMaxSegmentSize.

Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com>
---
 drivers/usb/gadget/function/f_ncm.c | 51 +++++++++++++++++++++++++++++
 drivers/usb/gadget/function/u_ncm.h |  2 ++
 2 files changed, 53 insertions(+)

Comments

Maciej Żenczykowski Oct. 10, 2023, 12:20 a.m. UTC | #1
On Mon, Oct 9, 2023 at 5:17 PM Maciej Żenczykowski <maze@google.com> wrote:
>
> On Mon, Oct 9, 2023 at 7:20 AM Krishna Kurapati
> <quic_kriskura@quicinc.com> wrote:
> >
> > Currently the NCM driver restricts wMaxSegmentSize that indicates
> > the datagram size coming from network layer to 1514. However the
> > spec doesn't have any limitation. For P2P connections over NCM,
> > increasing MTU helps increasing throughput.
> >
> > Add support to configure this value before configfs symlink is
> > created. Also since the NTB Out/In buffer sizes are fixed at 16384
> > bytes, limit the segment size to an upper cap of 15014. Set the
> > default MTU size for the ncm interface during function bind before
> > network interface is registered allowing MTU to be set in parity
> > with wMaxSegmentSize.
> >
> > Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com>
> > ---
> >  drivers/usb/gadget/function/f_ncm.c | 51 +++++++++++++++++++++++++++++
> >  drivers/usb/gadget/function/u_ncm.h |  2 ++
> >  2 files changed, 53 insertions(+)
> >
> > diff --git a/drivers/usb/gadget/function/f_ncm.c b/drivers/usb/gadget/function/f_ncm.c
> > index feccf4c8cc4f..eab297b22200 100644
> > --- a/drivers/usb/gadget/function/f_ncm.c
> > +++ b/drivers/usb/gadget/function/f_ncm.c
> > @@ -103,6 +103,8 @@ static inline struct f_ncm *func_to_ncm(struct usb_function *f)
> >  /* Delay for the transmit to wait before sending an unfilled NTB frame. */
> >  #define TX_TIMEOUT_NSECS       300000
> >
> > +#define MAX_DATAGRAM_SIZE      15014
> > +
> >  #define FORMATS_SUPPORTED      (USB_CDC_NCM_NTB16_SUPPORTED |  \
> >                                  USB_CDC_NCM_NTB32_SUPPORTED)
> >
> > @@ -1408,6 +1410,7 @@ static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
> >         ncm_opts = container_of(f->fi, struct f_ncm_opts, func_inst);
> >
> >         if (cdev->use_os_string) {
> > +               ncm_opts->net->mtu = (ncm_opts->max_segment_size - ETH_HLEN);
> >                 f->os_desc_table = kzalloc(sizeof(*f->os_desc_table),
> >                                            GFP_KERNEL);
> >                 if (!f->os_desc_table)
> > @@ -1469,6 +1472,8 @@ static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
> >
> >         status = -ENODEV;
> >
> > +       ecm_desc.wMaxSegmentSize = ncm_opts->max_segment_size;
>
> I think you need byte swap here.
>
> > +
> >         /* allocate instance-specific endpoints */
> >         ep = usb_ep_autoconfig(cdev->gadget, &fs_ncm_in_desc);
> >         if (!ep)
> > @@ -1569,11 +1574,56 @@ USB_ETHERNET_CONFIGFS_ITEM_ATTR_QMULT(ncm);
> >  /* f_ncm_opts_ifname */
> >  USB_ETHERNET_CONFIGFS_ITEM_ATTR_IFNAME(ncm);
> >
> > +static ssize_t ncm_opts_max_segment_size_show(struct config_item *item,
> > +                                             char *page)
> > +{
> > +       struct f_ncm_opts *opts = to_f_ncm_opts(item);
> > +       u32 segment_size;
> > +
> > +       mutex_lock(&opts->lock);
> > +       segment_size = opts->max_segment_size;
> > +       mutex_unlock(&opts->lock);
> > +
> > +       return sprintf(page, "%u\n", segment_size);
> > +}
> > +
> > +static ssize_t ncm_opts_max_segment_size_store(struct config_item *item,
> > +                                              const char *page, size_t len)
> > +{
> > +       struct f_ncm_opts *opts = to_f_ncm_opts(item);
> > +       int ret;
> > +       u32 segment_size;
> > +
> > +       mutex_lock(&opts->lock);
> > +       if (opts->refcnt) {
> > +               ret = -EBUSY;
> > +               goto out;
> > +       }
> > +
> > +       ret = kstrtou32(page, 0, &segment_size);
> > +       if (ret)
> > +               goto out;
> > +
> > +       if (segment_size > MAX_DATAGRAM_SIZE) {
> > +               ret = -EINVAL;
> > +               goto out;
> > +       }
> > +
> > +       opts->max_segment_size = segment_size;
> > +       ret = len;
> > +out:
> > +       mutex_unlock(&opts->lock);
> > +       return ret;
> > +}
> > +
> > +CONFIGFS_ATTR(ncm_opts_, max_segment_size);
> > +
> >  static struct configfs_attribute *ncm_attrs[] = {
> >         &ncm_opts_attr_dev_addr,
> >         &ncm_opts_attr_host_addr,
> >         &ncm_opts_attr_qmult,
> >         &ncm_opts_attr_ifname,
> > +       &ncm_opts_attr_max_segment_size,
> >         NULL,
> >  };
> >
> > @@ -1616,6 +1666,7 @@ static struct usb_function_instance *ncm_alloc_inst(void)
> >                 kfree(opts);
> >                 return ERR_CAST(net);
> >         }
> > +       opts->max_segment_size = cpu_to_le16(ETH_FRAME_LEN);
>
> and not here.  ie. max_segment_size should be native endian
>
> >         INIT_LIST_HEAD(&opts->ncm_os_desc.ext_prop);
> >
> >         descs[0] = &opts->ncm_os_desc;
> > diff --git a/drivers/usb/gadget/function/u_ncm.h b/drivers/usb/gadget/function/u_ncm.h
> > index 5408854d8407..d3403cf13f17 100644
> > --- a/drivers/usb/gadget/function/u_ncm.h
> > +++ b/drivers/usb/gadget/function/u_ncm.h
> > @@ -31,6 +31,8 @@ struct f_ncm_opts {
> >          */
> >         struct mutex                    lock;
> >         int                             refcnt;
> > +
> > +       u32                             max_segment_size;
> >  };
> >
> >  #endif /* U_NCM_H */
> > --
> > 2.42.0
> >
>
> That said, I don't really follow what this is doing...

Also

static struct usb_cdc_ether_desc ecm_desc = {
...
.wMaxSegmentSize = cpu_to_le16(ETH_FRAME_LEN),

^ I think this should be deleted now, right?  since it's always overwritten?
And if it isn't always overwritten, that would be a bug I think, cause
what happens if you bring up 2 ncm devices and only change the setting
on the 1st?
Krishna Kurapati Oct. 10, 2023, 4:34 a.m. UTC | #2
On 10/10/2023 5:50 AM, Maciej Żenczykowski wrote:
> On Mon, Oct 9, 2023 at 5:17 PM Maciej Żenczykowski <maze@google.com> wrote:
>>
>> On Mon, Oct 9, 2023 at 7:20 AM Krishna Kurapati
>> <quic_kriskura@quicinc.com> wrote:
>>>
>>> Currently the NCM driver restricts wMaxSegmentSize that indicates
>>> the datagram size coming from network layer to 1514. However the
>>> spec doesn't have any limitation. For P2P connections over NCM,
>>> increasing MTU helps increasing throughput.
>>>
>>> Add support to configure this value before configfs symlink is
>>> created. Also since the NTB Out/In buffer sizes are fixed at 16384
>>> bytes, limit the segment size to an upper cap of 15014. Set the
>>> default MTU size for the ncm interface during function bind before
>>> network interface is registered allowing MTU to be set in parity
>>> with wMaxSegmentSize.
>>>
>>> Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com>
>>> ---
>>>   drivers/usb/gadget/function/f_ncm.c | 51 +++++++++++++++++++++++++++++
>>>   drivers/usb/gadget/function/u_ncm.h |  2 ++
>>>   2 files changed, 53 insertions(+)
>>>
>>> diff --git a/drivers/usb/gadget/function/f_ncm.c b/drivers/usb/gadget/function/f_ncm.c
>>> index feccf4c8cc4f..eab297b22200 100644
>>> --- a/drivers/usb/gadget/function/f_ncm.c
>>> +++ b/drivers/usb/gadget/function/f_ncm.c
>>> @@ -103,6 +103,8 @@ static inline struct f_ncm *func_to_ncm(struct usb_function *f)
>>>   /* Delay for the transmit to wait before sending an unfilled NTB frame. */
>>>   #define TX_TIMEOUT_NSECS       300000
>>>
>>> +#define MAX_DATAGRAM_SIZE      15014
>>> +
>>>   #define FORMATS_SUPPORTED      (USB_CDC_NCM_NTB16_SUPPORTED |  \
>>>                                   USB_CDC_NCM_NTB32_SUPPORTED)
>>>
>>> @@ -1408,6 +1410,7 @@ static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
>>>          ncm_opts = container_of(f->fi, struct f_ncm_opts, func_inst);
>>>
>>>          if (cdev->use_os_string) {
>>> +               ncm_opts->net->mtu = (ncm_opts->max_segment_size - ETH_HLEN);
>>>                  f->os_desc_table = kzalloc(sizeof(*f->os_desc_table),
>>>                                             GFP_KERNEL);
>>>                  if (!f->os_desc_table)
>>> @@ -1469,6 +1472,8 @@ static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
>>>
>>>          status = -ENODEV;
>>>
>>> +       ecm_desc.wMaxSegmentSize = ncm_opts->max_segment_size;
>>
>> I think you need byte swap here.
>>
>>> +
>>>          /* allocate instance-specific endpoints */
>>>          ep = usb_ep_autoconfig(cdev->gadget, &fs_ncm_in_desc);
>>>          if (!ep)
>>> @@ -1569,11 +1574,56 @@ USB_ETHERNET_CONFIGFS_ITEM_ATTR_QMULT(ncm);
>>>   /* f_ncm_opts_ifname */
>>>   USB_ETHERNET_CONFIGFS_ITEM_ATTR_IFNAME(ncm);
>>>
>>> +static ssize_t ncm_opts_max_segment_size_show(struct config_item *item,
>>> +                                             char *page)
>>> +{
>>> +       struct f_ncm_opts *opts = to_f_ncm_opts(item);
>>> +       u32 segment_size;
>>> +
>>> +       mutex_lock(&opts->lock);
>>> +       segment_size = opts->max_segment_size;
>>> +       mutex_unlock(&opts->lock);
>>> +
>>> +       return sprintf(page, "%u\n", segment_size);
>>> +}
>>> +
>>> +static ssize_t ncm_opts_max_segment_size_store(struct config_item *item,
>>> +                                              const char *page, size_t len)
>>> +{
>>> +       struct f_ncm_opts *opts = to_f_ncm_opts(item);
>>> +       int ret;
>>> +       u32 segment_size;
>>> +
>>> +       mutex_lock(&opts->lock);
>>> +       if (opts->refcnt) {
>>> +               ret = -EBUSY;
>>> +               goto out;
>>> +       }
>>> +
>>> +       ret = kstrtou32(page, 0, &segment_size);
>>> +       if (ret)
>>> +               goto out;
>>> +
>>> +       if (segment_size > MAX_DATAGRAM_SIZE) {
>>> +               ret = -EINVAL;
>>> +               goto out;
>>> +       }
>>> +
>>> +       opts->max_segment_size = segment_size;
>>> +       ret = len;
>>> +out:
>>> +       mutex_unlock(&opts->lock);
>>> +       return ret;
>>> +}
>>> +
>>> +CONFIGFS_ATTR(ncm_opts_, max_segment_size);
>>> +
>>>   static struct configfs_attribute *ncm_attrs[] = {
>>>          &ncm_opts_attr_dev_addr,
>>>          &ncm_opts_attr_host_addr,
>>>          &ncm_opts_attr_qmult,
>>>          &ncm_opts_attr_ifname,
>>> +       &ncm_opts_attr_max_segment_size,
>>>          NULL,
>>>   };
>>>
>>> @@ -1616,6 +1666,7 @@ static struct usb_function_instance *ncm_alloc_inst(void)
>>>                  kfree(opts);
>>>                  return ERR_CAST(net);
>>>          }
>>> +       opts->max_segment_size = cpu_to_le16(ETH_FRAME_LEN);
>>
>> and not here.  ie. max_segment_size should be native endian
>>
>>>          INIT_LIST_HEAD(&opts->ncm_os_desc.ext_prop);
>>>
>>>          descs[0] = &opts->ncm_os_desc;
>>> diff --git a/drivers/usb/gadget/function/u_ncm.h b/drivers/usb/gadget/function/u_ncm.h
>>> index 5408854d8407..d3403cf13f17 100644
>>> --- a/drivers/usb/gadget/function/u_ncm.h
>>> +++ b/drivers/usb/gadget/function/u_ncm.h
>>> @@ -31,6 +31,8 @@ struct f_ncm_opts {
>>>           */
>>>          struct mutex                    lock;
>>>          int                             refcnt;
>>> +
>>> +       u32                             max_segment_size;
>>>   };
>>>
>>>   #endif /* U_NCM_H */
>>> --
>>> 2.42.0
>>>
>>
>> That said, I don't really follow what this is doing...
> 
> Also
> 
> static struct usb_cdc_ether_desc ecm_desc = {
> ...
> .wMaxSegmentSize = cpu_to_le16(ETH_FRAME_LEN),
> 
> ^ I think this should be deleted now, right?  since it's always overwritten?
> And if it isn't always overwritten, that would be a bug I think, cause
> what happens if you bring up 2 ncm devices and only change the setting
> on the 1st?

Actually, keeping it or removing it doesn't affect because the default 
value of this will be set in bind call via opts->max_segment_size. And 
this max_segment_size variable is defaulted to ETH_FRAME_LEN in 
alloc_inst. After bind, writing to this value doesn't take effect. 
Before the netdev interface was registered, if this value is updated 
then MTU is changed automatically for this interface. If there are two 
NCM devices, then each of them will have their own instance of 
wMaxSegmentSize property in their usb_gadget/g1/functions/ncm.X 
directories and both of them are defaulted ETH_FRAME_LEN if the configfs 
property is not disturbed. In case only one needs to change, we can 
change via configfs and the other stays as ETH_FRAME_LEN only.

Regards,
Krishna,
kernel test robot Oct. 10, 2023, 10:26 p.m. UTC | #3
Hi Krishna,

kernel test robot noticed the following build warnings:

[auto build test WARNING on usb/usb-testing]
[also build test WARNING on usb/usb-next usb/usb-linus linus/master v6.6-rc5 next-20231010]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Krishna-Kurapati/usb-gadget-ncm-Add-support-to-update-wMaxSegmentSize-via-configfs/20231009-222315
base:   https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git usb-testing
patch link:    https://lore.kernel.org/r/20231009142005.21338-2-quic_kriskura%40quicinc.com
patch subject: [PATCH 2/2] usb: gadget: ncm: Add support to update wMaxSegmentSize via configfs
config: i386-randconfig-062-20231010 (https://download.01.org/0day-ci/archive/20231011/202310110658.n9yg3tJy-lkp@intel.com/config)
compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20231011/202310110658.n9yg3tJy-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202310110658.n9yg3tJy-lkp@intel.com/

sparse warnings: (new ones prefixed by >>)
>> drivers/usb/gadget/function/f_ncm.c:1475:34: sparse: sparse: incorrect type in assignment (different base types) @@     expected restricted __le16 static [addressable] [assigned] [toplevel] [usertype] wMaxSegmentSize @@     got unsigned int [usertype] max_segment_size @@
   drivers/usb/gadget/function/f_ncm.c:1475:34: sparse:     expected restricted __le16 static [addressable] [assigned] [toplevel] [usertype] wMaxSegmentSize
   drivers/usb/gadget/function/f_ncm.c:1475:34: sparse:     got unsigned int [usertype] max_segment_size
>> drivers/usb/gadget/function/f_ncm.c:1669:32: sparse: sparse: incorrect type in assignment (different base types) @@     expected unsigned int [usertype] max_segment_size @@     got restricted __le16 [usertype] @@
   drivers/usb/gadget/function/f_ncm.c:1669:32: sparse:     expected unsigned int [usertype] max_segment_size
   drivers/usb/gadget/function/f_ncm.c:1669:32: sparse:     got restricted __le16 [usertype]

vim +1475 drivers/usb/gadget/function/f_ncm.c

  1397	
  1398	static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
  1399	{
  1400		struct usb_composite_dev *cdev = c->cdev;
  1401		struct f_ncm		*ncm = func_to_ncm(f);
  1402		struct usb_string	*us;
  1403		int			status;
  1404		struct usb_ep		*ep;
  1405		struct f_ncm_opts	*ncm_opts;
  1406	
  1407		if (!can_support_ecm(cdev->gadget))
  1408			return -EINVAL;
  1409	
  1410		ncm_opts = container_of(f->fi, struct f_ncm_opts, func_inst);
  1411	
  1412		if (cdev->use_os_string) {
  1413			ncm_opts->net->mtu = (ncm_opts->max_segment_size - ETH_HLEN);
  1414			f->os_desc_table = kzalloc(sizeof(*f->os_desc_table),
  1415						   GFP_KERNEL);
  1416			if (!f->os_desc_table)
  1417				return -ENOMEM;
  1418			f->os_desc_n = 1;
  1419			f->os_desc_table[0].os_desc = &ncm_opts->ncm_os_desc;
  1420		}
  1421	
  1422		/*
  1423		 * in drivers/usb/gadget/configfs.c:configfs_composite_bind()
  1424		 * configurations are bound in sequence with list_for_each_entry,
  1425		 * in each configuration its functions are bound in sequence
  1426		 * with list_for_each_entry, so we assume no race condition
  1427		 * with regard to ncm_opts->bound access
  1428		 */
  1429		if (!ncm_opts->bound) {
  1430			mutex_lock(&ncm_opts->lock);
  1431			gether_set_gadget(ncm_opts->net, cdev->gadget);
  1432			status = gether_register_netdev(ncm_opts->net);
  1433			mutex_unlock(&ncm_opts->lock);
  1434			if (status)
  1435				goto fail;
  1436			ncm_opts->bound = true;
  1437		}
  1438		us = usb_gstrings_attach(cdev, ncm_strings,
  1439					 ARRAY_SIZE(ncm_string_defs));
  1440		if (IS_ERR(us)) {
  1441			status = PTR_ERR(us);
  1442			goto fail;
  1443		}
  1444		ncm_control_intf.iInterface = us[STRING_CTRL_IDX].id;
  1445		ncm_data_nop_intf.iInterface = us[STRING_DATA_IDX].id;
  1446		ncm_data_intf.iInterface = us[STRING_DATA_IDX].id;
  1447		ecm_desc.iMACAddress = us[STRING_MAC_IDX].id;
  1448		ncm_iad_desc.iFunction = us[STRING_IAD_IDX].id;
  1449	
  1450		/* allocate instance-specific interface IDs */
  1451		status = usb_interface_id(c, f);
  1452		if (status < 0)
  1453			goto fail;
  1454		ncm->ctrl_id = status;
  1455		ncm_iad_desc.bFirstInterface = status;
  1456	
  1457		ncm_control_intf.bInterfaceNumber = status;
  1458		ncm_union_desc.bMasterInterface0 = status;
  1459	
  1460		if (cdev->use_os_string)
  1461			f->os_desc_table[0].if_id =
  1462				ncm_iad_desc.bFirstInterface;
  1463	
  1464		status = usb_interface_id(c, f);
  1465		if (status < 0)
  1466			goto fail;
  1467		ncm->data_id = status;
  1468	
  1469		ncm_data_nop_intf.bInterfaceNumber = status;
  1470		ncm_data_intf.bInterfaceNumber = status;
  1471		ncm_union_desc.bSlaveInterface0 = status;
  1472	
  1473		status = -ENODEV;
  1474	
> 1475		ecm_desc.wMaxSegmentSize = ncm_opts->max_segment_size;
  1476	
  1477		/* allocate instance-specific endpoints */
  1478		ep = usb_ep_autoconfig(cdev->gadget, &fs_ncm_in_desc);
  1479		if (!ep)
  1480			goto fail;
  1481		ncm->port.in_ep = ep;
  1482	
  1483		ep = usb_ep_autoconfig(cdev->gadget, &fs_ncm_out_desc);
  1484		if (!ep)
  1485			goto fail;
  1486		ncm->port.out_ep = ep;
  1487	
  1488		ep = usb_ep_autoconfig(cdev->gadget, &fs_ncm_notify_desc);
  1489		if (!ep)
  1490			goto fail;
  1491		ncm->notify = ep;
  1492	
  1493		status = -ENOMEM;
  1494	
  1495		/* allocate notification request and buffer */
  1496		ncm->notify_req = usb_ep_alloc_request(ep, GFP_KERNEL);
  1497		if (!ncm->notify_req)
  1498			goto fail;
  1499		ncm->notify_req->buf = kmalloc(NCM_STATUS_BYTECOUNT, GFP_KERNEL);
  1500		if (!ncm->notify_req->buf)
  1501			goto fail;
  1502		ncm->notify_req->context = ncm;
  1503		ncm->notify_req->complete = ncm_notify_complete;
  1504	
  1505		/*
  1506		 * support all relevant hardware speeds... we expect that when
  1507		 * hardware is dual speed, all bulk-capable endpoints work at
  1508		 * both speeds
  1509		 */
  1510		hs_ncm_in_desc.bEndpointAddress = fs_ncm_in_desc.bEndpointAddress;
  1511		hs_ncm_out_desc.bEndpointAddress = fs_ncm_out_desc.bEndpointAddress;
  1512		hs_ncm_notify_desc.bEndpointAddress =
  1513			fs_ncm_notify_desc.bEndpointAddress;
  1514	
  1515		ss_ncm_in_desc.bEndpointAddress = fs_ncm_in_desc.bEndpointAddress;
  1516		ss_ncm_out_desc.bEndpointAddress = fs_ncm_out_desc.bEndpointAddress;
  1517		ss_ncm_notify_desc.bEndpointAddress =
  1518			fs_ncm_notify_desc.bEndpointAddress;
  1519	
  1520		status = usb_assign_descriptors(f, ncm_fs_function, ncm_hs_function,
  1521				ncm_ss_function, ncm_ss_function);
  1522		if (status)
  1523			goto fail;
  1524	
  1525		/*
  1526		 * NOTE:  all that is done without knowing or caring about
  1527		 * the network link ... which is unavailable to this code
  1528		 * until we're activated via set_alt().
  1529		 */
  1530	
  1531		ncm->port.open = ncm_open;
  1532		ncm->port.close = ncm_close;
  1533	
  1534		hrtimer_init(&ncm->task_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_SOFT);
  1535		ncm->task_timer.function = ncm_tx_timeout;
  1536	
  1537		DBG(cdev, "CDC Network: IN/%s OUT/%s NOTIFY/%s\n",
  1538				ncm->port.in_ep->name, ncm->port.out_ep->name,
  1539				ncm->notify->name);
  1540		return 0;
  1541	
  1542	fail:
  1543		kfree(f->os_desc_table);
  1544		f->os_desc_n = 0;
  1545	
  1546		if (ncm->notify_req) {
  1547			kfree(ncm->notify_req->buf);
  1548			usb_ep_free_request(ncm->notify, ncm->notify_req);
  1549		}
  1550	
  1551		ERROR(cdev, "%s: can't bind, err %d\n", f->name, status);
  1552	
  1553		return status;
  1554	}
  1555	
  1556	static inline struct f_ncm_opts *to_f_ncm_opts(struct config_item *item)
  1557	{
  1558		return container_of(to_config_group(item), struct f_ncm_opts,
  1559				    func_inst.group);
  1560	}
  1561	
  1562	/* f_ncm_item_ops */
  1563	USB_ETHERNET_CONFIGFS_ITEM(ncm);
  1564	
  1565	/* f_ncm_opts_dev_addr */
  1566	USB_ETHERNET_CONFIGFS_ITEM_ATTR_DEV_ADDR(ncm);
  1567	
  1568	/* f_ncm_opts_host_addr */
  1569	USB_ETHERNET_CONFIGFS_ITEM_ATTR_HOST_ADDR(ncm);
  1570	
  1571	/* f_ncm_opts_qmult */
  1572	USB_ETHERNET_CONFIGFS_ITEM_ATTR_QMULT(ncm);
  1573	
  1574	/* f_ncm_opts_ifname */
  1575	USB_ETHERNET_CONFIGFS_ITEM_ATTR_IFNAME(ncm);
  1576	
  1577	static ssize_t ncm_opts_max_segment_size_show(struct config_item *item,
  1578						      char *page)
  1579	{
  1580		struct f_ncm_opts *opts = to_f_ncm_opts(item);
  1581		u32 segment_size;
  1582	
  1583		mutex_lock(&opts->lock);
  1584		segment_size = opts->max_segment_size;
  1585		mutex_unlock(&opts->lock);
  1586	
  1587		return sprintf(page, "%u\n", segment_size);
  1588	}
  1589	
  1590	static ssize_t ncm_opts_max_segment_size_store(struct config_item *item,
  1591						       const char *page, size_t len)
  1592	{
  1593		struct f_ncm_opts *opts = to_f_ncm_opts(item);
  1594		int ret;
  1595		u32 segment_size;
  1596	
  1597		mutex_lock(&opts->lock);
  1598		if (opts->refcnt) {
  1599			ret = -EBUSY;
  1600			goto out;
  1601		}
  1602	
  1603		ret = kstrtou32(page, 0, &segment_size);
  1604		if (ret)
  1605			goto out;
  1606	
  1607		if (segment_size > MAX_DATAGRAM_SIZE) {
  1608			ret = -EINVAL;
  1609			goto out;
  1610		}
  1611	
  1612		opts->max_segment_size = segment_size;
  1613		ret = len;
  1614	out:
  1615		mutex_unlock(&opts->lock);
  1616		return ret;
  1617	}
  1618	
  1619	CONFIGFS_ATTR(ncm_opts_, max_segment_size);
  1620	
  1621	static struct configfs_attribute *ncm_attrs[] = {
  1622		&ncm_opts_attr_dev_addr,
  1623		&ncm_opts_attr_host_addr,
  1624		&ncm_opts_attr_qmult,
  1625		&ncm_opts_attr_ifname,
  1626		&ncm_opts_attr_max_segment_size,
  1627		NULL,
  1628	};
  1629	
  1630	static const struct config_item_type ncm_func_type = {
  1631		.ct_item_ops	= &ncm_item_ops,
  1632		.ct_attrs	= ncm_attrs,
  1633		.ct_owner	= THIS_MODULE,
  1634	};
  1635	
  1636	static void ncm_free_inst(struct usb_function_instance *f)
  1637	{
  1638		struct f_ncm_opts *opts;
  1639	
  1640		opts = container_of(f, struct f_ncm_opts, func_inst);
  1641		if (opts->bound)
  1642			gether_cleanup(netdev_priv(opts->net));
  1643		else
  1644			free_netdev(opts->net);
  1645		kfree(opts->ncm_interf_group);
  1646		kfree(opts);
  1647	}
  1648	
  1649	static struct usb_function_instance *ncm_alloc_inst(void)
  1650	{
  1651		struct f_ncm_opts *opts;
  1652		struct usb_os_desc *descs[1];
  1653		char *names[1];
  1654		struct config_group *ncm_interf_group;
  1655	
  1656		opts = kzalloc(sizeof(*opts), GFP_KERNEL);
  1657		if (!opts)
  1658			return ERR_PTR(-ENOMEM);
  1659		opts->ncm_os_desc.ext_compat_id = opts->ncm_ext_compat_id;
  1660	
  1661		mutex_init(&opts->lock);
  1662		opts->func_inst.free_func_inst = ncm_free_inst;
  1663		opts->net = gether_setup_default();
  1664		if (IS_ERR(opts->net)) {
  1665			struct net_device *net = opts->net;
  1666			kfree(opts);
  1667			return ERR_CAST(net);
  1668		}
> 1669		opts->max_segment_size = cpu_to_le16(ETH_FRAME_LEN);
  1670		INIT_LIST_HEAD(&opts->ncm_os_desc.ext_prop);
  1671	
  1672		descs[0] = &opts->ncm_os_desc;
  1673		names[0] = "ncm";
  1674	
  1675		config_group_init_type_name(&opts->func_inst.group, "", &ncm_func_type);
  1676		ncm_interf_group =
  1677			usb_os_desc_prepare_interf_dir(&opts->func_inst.group, 1, descs,
  1678						       names, THIS_MODULE);
  1679		if (IS_ERR(ncm_interf_group)) {
  1680			ncm_free_inst(&opts->func_inst);
  1681			return ERR_CAST(ncm_interf_group);
  1682		}
  1683		opts->ncm_interf_group = ncm_interf_group;
  1684	
  1685		return &opts->func_inst;
  1686	}
  1687
Krishna Kurapati Oct. 12, 2023, 8:48 a.m. UTC | #4
On 10/10/2023 10:08 AM, Krishna Kurapati PSSNV wrote:
> 

>>
>> ^ is this a problem now if we have >1 gadget?
>> how does it work then?
> 
> 
> You are right. This would effect unwrap call and the wMaxSegmentSize is 
> used directly. Thanks for the catch. I didn't test with 2 NCM interfaces 
> and hence I wasn't able to find this bug. Perhaps changing this to 
> opts->max_segment_size would fix the implementation as unwrap would 
> anyways be called after bind.

Hi Maciej,

  How about the below diff:

---------

+/*
+ * Allow max segment size to be in parity with max_mtu possible
+ * for the interface.
+ */
+#define MAX_DATAGRAM_SIZE      GETHER_MAX_ETH_FRAME_LEN
+
  #define FORMATS_SUPPORTED      (USB_CDC_NCM_NTB16_SUPPORTED |  \
                                  USB_CDC_NCM_NTB32_SUPPORTED)

@@ -194,7 +200,6 @@ static struct usb_cdc_ether_desc ecm_desc = {
         /* this descriptor actually adds value, surprise! */
         /* .iMACAddress = DYNAMIC */
         .bmEthernetStatistics = cpu_to_le32(0), /* no statistics */
-       .wMaxSegmentSize =      cpu_to_le16(ETH_FRAME_LEN),
         .wNumberMCFilters =     cpu_to_le16(0),
         .bNumberPowerFilters =  0,
  };
@@ -1180,10 +1185,15 @@ static int ncm_unwrap_ntb(struct gether *port,
         struct sk_buff  *skb2;
         int             ret = -EINVAL;
         unsigned        ntb_max = 
le32_to_cpu(ntb_parameters.dwNtbOutMaxSize);
-       unsigned        frame_max = le16_to_cpu(ecm_desc.wMaxSegmentSize);
+       unsigned int    frame_max;
         const struct ndp_parser_opts *opts = ncm->parser_opts;
         unsigned        crc_len = ncm->is_crc ? sizeof(uint32_t) : 0;
         int             dgram_counter;
+       struct f_ncm_opts *ncm_opts;
+       const struct usb_function_instance *fi = port->func.fi;
+
+       ncm_opts = container_of(fi, struct f_ncm_opts, func_inst);
+       frame_max = ncm_opts->max_segment_size;

         /* dwSignature */
         if (get_unaligned_le32(tmp) != opts->nth_sign) {
@@ -1440,6 +1450,7 @@ static int ncm_bind(struct usb_configuration *c, 
struct usb_function *f)
          */
         if (!ncm_opts->bound) {
                 mutex_lock(&ncm_opts->lock);
+               ncm_opts->net->mtu = (ncm_opts->max_segment_size - 
ETH_HLEN);
                 gether_set_gadget(ncm_opts->net, cdev->gadget);
                 status = gether_register_netdev(ncm_opts->net);
                 mutex_unlock(&ncm_opts->lock);
@@ -1484,6 +1495,8 @@ static int ncm_bind(struct usb_configuration *c, 
struct usb_function *f)

         status = -ENODEV;

+       ecm_desc.wMaxSegmentSize = (__le16)ncm_opts->max_segment_size;
+

------

I can limit the max segment size to (Max MTU + ETH_HELN) and this would 
be logical to do. Also we can set the frame_max from ncm_opts itself 
while initializing it to 1514 (default value) during alloc_inst callback 
and nothing would break while still being backward compatible.

Let me know your thoughts on this.

Regards,
Krishna,
Maciej Żenczykowski Oct. 12, 2023, 12:32 p.m. UTC | #5
On Thu, Oct 12, 2023 at 1:48 AM Krishna Kurapati PSSNV
<quic_kriskura@quicinc.com> wrote:
>
>
>
> On 10/10/2023 10:08 AM, Krishna Kurapati PSSNV wrote:
> >
>
> >>
> >> ^ is this a problem now if we have >1 gadget?
> >> how does it work then?
> >
> >
> > You are right. This would effect unwrap call and the wMaxSegmentSize is
> > used directly. Thanks for the catch. I didn't test with 2 NCM interfaces
> > and hence I wasn't able to find this bug. Perhaps changing this to
> > opts->max_segment_size would fix the implementation as unwrap would
> > anyways be called after bind.
>
> Hi Maciej,
>
>   How about the below diff:
>
> ---------
>
> +/*
> + * Allow max segment size to be in parity with max_mtu possible
> + * for the interface.
> + */
> +#define MAX_DATAGRAM_SIZE      GETHER_MAX_ETH_FRAME_LEN
> +
>   #define FORMATS_SUPPORTED      (USB_CDC_NCM_NTB16_SUPPORTED |  \
>                                   USB_CDC_NCM_NTB32_SUPPORTED)
>
> @@ -194,7 +200,6 @@ static struct usb_cdc_ether_desc ecm_desc = {
>          /* this descriptor actually adds value, surprise! */
>          /* .iMACAddress = DYNAMIC */
>          .bmEthernetStatistics = cpu_to_le32(0), /* no statistics */
> -       .wMaxSegmentSize =      cpu_to_le16(ETH_FRAME_LEN),
>          .wNumberMCFilters =     cpu_to_le16(0),
>          .bNumberPowerFilters =  0,
>   };
> @@ -1180,10 +1185,15 @@ static int ncm_unwrap_ntb(struct gether *port,
>          struct sk_buff  *skb2;
>          int             ret = -EINVAL;
>          unsigned        ntb_max =
> le32_to_cpu(ntb_parameters.dwNtbOutMaxSize);
> -       unsigned        frame_max = le16_to_cpu(ecm_desc.wMaxSegmentSize);
> +       unsigned int    frame_max;
>          const struct ndp_parser_opts *opts = ncm->parser_opts;
>          unsigned        crc_len = ncm->is_crc ? sizeof(uint32_t) : 0;
>          int             dgram_counter;
> +       struct f_ncm_opts *ncm_opts;
> +       const struct usb_function_instance *fi = port->func.fi;
> +
> +       ncm_opts = container_of(fi, struct f_ncm_opts, func_inst);
> +       frame_max = ncm_opts->max_segment_size;
>
>          /* dwSignature */
>          if (get_unaligned_le32(tmp) != opts->nth_sign) {
> @@ -1440,6 +1450,7 @@ static int ncm_bind(struct usb_configuration *c,
> struct usb_function *f)
>           */
>          if (!ncm_opts->bound) {
>                  mutex_lock(&ncm_opts->lock);
> +               ncm_opts->net->mtu = (ncm_opts->max_segment_size -
> ETH_HLEN);
>                  gether_set_gadget(ncm_opts->net, cdev->gadget);
>                  status = gether_register_netdev(ncm_opts->net);
>                  mutex_unlock(&ncm_opts->lock);
> @@ -1484,6 +1495,8 @@ static int ncm_bind(struct usb_configuration *c,
> struct usb_function *f)
>
>          status = -ENODEV;
>
> +       ecm_desc.wMaxSegmentSize = (__le16)ncm_opts->max_segment_size;

this looks wrong. pretty sure this should be some form of cpu_to_le16

> +
>
> ------
>
> I can limit the max segment size to (Max MTU + ETH_HELN) and this would
> be logical to do. Also we can set the frame_max from ncm_opts itself
> while initializing it to 1514 (default value) during alloc_inst callback
> and nothing would break while still being backward compatible.
>
> Let me know your thoughts on this.
>
> Regards,
> Krishna,

Could you paste the full patch?
This is hard to review without looking at much more context then email
is providing
(or, even better, send me a link to a CL in gerrit somewhere - for
example aosp ACK mainline tree)
Maciej Żenczykowski Oct. 13, 2023, 6:39 p.m. UTC | #6
On Thu, Oct 12, 2023 at 8:40 AM Krishna Kurapati PSSNV
<quic_kriskura@quicinc.com> wrote:
>
>
>
> On 10/12/2023 6:02 PM, Maciej Żenczykowski wrote:
> > On Thu, Oct 12, 2023 at 1:48 AM Krishna Kurapati PSSNV
> >
> > Could you paste the full patch?
> > This is hard to review without looking at much more context then email
> > is providing
> > (or, even better, send me a link to a CL in gerrit somewhere - for
> > example aosp ACK mainline tree)
>
> Sure. Will provide a gerrit on ACK for review before posting v2.
>
> The intent of posting the diff was two fold:
>
> 1. The question Greg asked regarding why the max segment size was
> limited to 15014 was valid. When I thought about it, I actually wanted
> to limit the max MTU to 15000, so the max segment size automatically
> needs to be limited to 15014.

Note that this is a *very* abstract value.
I get you want L3 MTU of 10 * 1500, but this value is not actually meaningful.

IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
do not result in a trivial multiplication of the standard 1500 byte
ethernet L3 MTU.
Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
frames depending on which type of aggregation you do.
(and for tcp it even depends on the number and size of tcp options,
though it is often assumed that those take up 12 bytes, since that's the
normal for Linux-to-Linux tcp connections)

For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu frames,
this means you have
N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
payload (1500-12-20-40=1500-72=1428)
post aggregation:
1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
payload (N*1428)

so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)

That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
it's 40/60 if there's no tcp options (which I think happens when
talking to windows)
it's different still with ipv4 fragmentation... and again different
with ipv6 fragmentation...
etc.

ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
Either way you don't get full frames.

As such I'd recommend going with whatever is the largest mtu that can
be meaningfully made to fit in 16K with all the NCM header overhead.
That's likely closer to 15500-16000 (though I have *not* checked).

> But my commit text didn't mention this
> properly which was a mistake on my behalf. But when I looked at the
> code, limiting the max segment size 15014 would force the practical
> max_mtu to not cross 15000 although theoretical max_mtu was set to:
> (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
>
> So my assumption of limiting it to 15000 was wrong. It must be limited
> to 15412 as mentioned in u_ether.c  This inturn means we must limit
> max_segment_size to:
> GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
> as mentioned in u_ether.c.
>
> I wanted to confirm that setting MAX_DATAGRAM_SIZE to
> GETHER_MAX_ETH_FRAME_LEN was correct.
>
> 2. I am not actually able to test with MTU beyond 15000. When my host
> device is a linux machine, the cdc_ncm.c limits max_segment_size to:
> CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */

In practice you get 50% of the benefits of infinitely large mtu by
going from 1500 to ~2980.
you get 75% of the benefits by going to ~6K
you get 87.5% of the benefits by going to ~12K
the benefits of going even higher are smaller and smaller...

If the host side is limited to 8192, maybe we should match that here too?

But the host side limitation of 8192 doesn't seem particularly sane either...
Maybe we should relax that instead?

(especially since for things like tcp zero copy you want an mtu which
is slighly more then N * 4096,
ie. around 4.5KB, 8.5KB, 12.5KB or something like that)

> When connected to windows machine, I am able to set the mtu to a max
> value of 15000. So not sure how to test the patch if I set the
> max_segment_size to GETHER_MAX_ETH_FRAME_LEN.
>
> By pasting the diff, I wanted to confirm both the above queries.
>
> And you are right, while assigning value to ecm.wMaxSegmentSize, we must
> use cpu_to_le16(...). Will ensure to make this change in v2. It worked
> without that too, not sure how.
Maciej Żenczykowski Oct. 13, 2023, 6:40 p.m. UTC | #7
On Fri, Oct 13, 2023 at 11:39 AM Maciej Żenczykowski <maze@google.com> wrote:
>
> On Thu, Oct 12, 2023 at 8:40 AM Krishna Kurapati PSSNV
> <quic_kriskura@quicinc.com> wrote:
> >
> >
> >
> > On 10/12/2023 6:02 PM, Maciej Żenczykowski wrote:
> > > On Thu, Oct 12, 2023 at 1:48 AM Krishna Kurapati PSSNV
> > >
> > > Could you paste the full patch?
> > > This is hard to review without looking at much more context then email
> > > is providing
> > > (or, even better, send me a link to a CL in gerrit somewhere - for
> > > example aosp ACK mainline tree)
> >
> > Sure. Will provide a gerrit on ACK for review before posting v2.
> >
> > The intent of posting the diff was two fold:
> >
> > 1. The question Greg asked regarding why the max segment size was
> > limited to 15014 was valid. When I thought about it, I actually wanted
> > to limit the max MTU to 15000, so the max segment size automatically
> > needs to be limited to 15014.
>
> Note that this is a *very* abstract value.
> I get you want L3 MTU of 10 * 1500, but this value is not actually meaningful.
>
> IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
> do not result in a trivial multiplication of the standard 1500 byte
> ethernet L3 MTU.
> Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
> frames depending on which type of aggregation you do.
> (and for tcp it even depends on the number and size of tcp options,
> though it is often assumed that those take up 12 bytes, since that's the
> normal for Linux-to-Linux tcp connections)
>
> For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu frames,
> this means you have
> N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> payload (1500-12-20-40=1500-72=1428)
> post aggregation:
> 1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> payload (N*1428)
>
> so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)

As you can see, for N=10, this isn't 15000, it's 72+10*1428 = 14352

>
> That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
> it's 40/60 if there's no tcp options (which I think happens when
> talking to windows)
> it's different still with ipv4 fragmentation... and again different
> with ipv6 fragmentation...
> etc.
>
> ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
> Either way you don't get full frames.
>
> As such I'd recommend going with whatever is the largest mtu that can
> be meaningfully made to fit in 16K with all the NCM header overhead.
> That's likely closer to 15500-16000 (though I have *not* checked).
>
> > But my commit text didn't mention this
> > properly which was a mistake on my behalf. But when I looked at the
> > code, limiting the max segment size 15014 would force the practical
> > max_mtu to not cross 15000 although theoretical max_mtu was set to:
> > (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
> >
> > So my assumption of limiting it to 15000 was wrong. It must be limited
> > to 15412 as mentioned in u_ether.c  This inturn means we must limit
> > max_segment_size to:
> > GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
> > as mentioned in u_ether.c.
> >
> > I wanted to confirm that setting MAX_DATAGRAM_SIZE to
> > GETHER_MAX_ETH_FRAME_LEN was correct.
> >
> > 2. I am not actually able to test with MTU beyond 15000. When my host
> > device is a linux machine, the cdc_ncm.c limits max_segment_size to:
> > CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */
>
> In practice you get 50% of the benefits of infinitely large mtu by
> going from 1500 to ~2980.
> you get 75% of the benefits by going to ~6K
> you get 87.5% of the benefits by going to ~12K
> the benefits of going even higher are smaller and smaller...
>
> If the host side is limited to 8192, maybe we should match that here too?
>
> But the host side limitation of 8192 doesn't seem particularly sane either...
> Maybe we should relax that instead?
>
> (especially since for things like tcp zero copy you want an mtu which
> is slighly more then N * 4096,
> ie. around 4.5KB, 8.5KB, 12.5KB or something like that)
>
> > When connected to windows machine, I am able to set the mtu to a max
> > value of 15000. So not sure how to test the patch if I set the
> > max_segment_size to GETHER_MAX_ETH_FRAME_LEN.
> >
> > By pasting the diff, I wanted to confirm both the above queries.
> >
> > And you are right, while assigning value to ecm.wMaxSegmentSize, we must
> > use cpu_to_le16(...). Will ensure to make this change in v2. It worked
> > without that too, not sure how.Maciej Żenczykowski, Kernel Networking Developer @ Google
Krishna Kurapati Oct. 13, 2023, 7:58 p.m. UTC | #8
On 10/14/2023 12:09 AM, Maciej Żenczykowski wrote:
> On Thu, Oct 12, 2023 at 8:40 AM Krishna Kurapati PSSNV
> <quic_kriskura@quicinc.com> wrote:
>>
>>
>>
>> On 10/12/2023 6:02 PM, Maciej Żenczykowski wrote:
>>> On Thu, Oct 12, 2023 at 1:48 AM Krishna Kurapati PSSNV
>>>
>>> Could you paste the full patch?
>>> This is hard to review without looking at much more context then email
>>> is providing
>>> (or, even better, send me a link to a CL in gerrit somewhere - for
>>> example aosp ACK mainline tree)
>>
>> Sure. Will provide a gerrit on ACK for review before posting v2.
>>
>> The intent of posting the diff was two fold:
>>
>> 1. The question Greg asked regarding why the max segment size was
>> limited to 15014 was valid. When I thought about it, I actually wanted
>> to limit the max MTU to 15000, so the max segment size automatically
>> needs to be limited to 15014.
> 
> Note that this is a *very* abstract value.
> I get you want L3 MTU of 10 * 1500, but this value is not actually meaningful.
> 
> IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
> do not result in a trivial multiplication of the standard 1500 byte
> ethernet L3 MTU.
> Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
> frames depending on which type of aggregation you do.
> (and for tcp it even depends on the number and size of tcp options,
> though it is often assumed that those take up 12 bytes, since that's the
> normal for Linux-to-Linux tcp connections)
> 
> For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu frames,
> this means you have
> N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> payload (1500-12-20-40=1500-72=1428)
> post aggregation:
> 1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> payload (N*1428)
> 
> so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)
> 
> That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
> it's 40/60 if there's no tcp options (which I think happens when
> talking to windows)
> it's different still with ipv4 fragmentation... and again different
> with ipv6 fragmentation...
> etc.
> 
> ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
> Either way you don't get full frames.
> 
> As such I'd recommend going with whatever is the largest mtu that can
> be meaningfully made to fit in 16K with all the NCM header overhead.
> That's likely closer to 15500-16000 (though I have *not* checked).
> 
>> But my commit text didn't mention this
>> properly which was a mistake on my behalf. But when I looked at the
>> code, limiting the max segment size 15014 would force the practical
>> max_mtu to not cross 15000 although theoretical max_mtu was set to:
>> (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
>>
>> So my assumption of limiting it to 15000 was wrong. It must be limited
>> to 15412 as mentioned in u_ether.c  This inturn means we must limit
>> max_segment_size to:
>> GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
>> as mentioned in u_ether.c.
>>
>> I wanted to confirm that setting MAX_DATAGRAM_SIZE to
>> GETHER_MAX_ETH_FRAME_LEN was correct.
>>
>> 2. I am not actually able to test with MTU beyond 15000. When my host
>> device is a linux machine, the cdc_ncm.c limits max_segment_size to:
>> CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */
> 
> In practice you get 50% of the benefits of infinitely large mtu by
> going from 1500 to ~2980.
> you get 75% of the benefits by going to ~6K
> you get 87.5% of the benefits by going to ~12K
> the benefits of going even higher are smaller and smaller...
>  > If the host side is limited to 8192, maybe we should match that here too?

Hi Maciej,

  Thanks for the detailed explanation. I agree with you on setting 
device side also to 8192 instead of what max_mtu is present in u_ether 
or practical max segment size possible.

> 
> But the host side limitation of 8192 doesn't seem particularly sane either...
> Maybe we should relax that instead?
> 
I really didn't understand why it was set to 8192 in first place.

> (especially since for things like tcp zero copy you want an mtu which
> is slighly more then N * 4096,
> ie. around 4.5KB, 8.5KB, 12.5KB or something like that)
> 

I am not sure about host mode completely. If we want to increase though, 
just increasing the MAX_DATAGRAM_SIZE to some bigger value help ? (I 
don't know the entire code of cdc_ncm, so I might be wrong).

Regards,
Krishna,
Maciej Żenczykowski Oct. 13, 2023, 10:35 p.m. UTC | #9
On Fri, Oct 13, 2023 at 12:58 PM Krishna Kurapati PSSNV
<quic_kriskura@quicinc.com> wrote:
>
>
>
> On 10/14/2023 12:09 AM, Maciej Żenczykowski wrote:
> > On Thu, Oct 12, 2023 at 8:40 AM Krishna Kurapati PSSNV
> > <quic_kriskura@quicinc.com> wrote:
> >>
> >>
> >>
> >> On 10/12/2023 6:02 PM, Maciej Żenczykowski wrote:
> >>> On Thu, Oct 12, 2023 at 1:48 AM Krishna Kurapati PSSNV
> >>>
> >>> Could you paste the full patch?
> >>> This is hard to review without looking at much more context then email
> >>> is providing
> >>> (or, even better, send me a link to a CL in gerrit somewhere - for
> >>> example aosp ACK mainline tree)
> >>
> >> Sure. Will provide a gerrit on ACK for review before posting v2.
> >>
> >> The intent of posting the diff was two fold:
> >>
> >> 1. The question Greg asked regarding why the max segment size was
> >> limited to 15014 was valid. When I thought about it, I actually wanted
> >> to limit the max MTU to 15000, so the max segment size automatically
> >> needs to be limited to 15014.
> >
> > Note that this is a *very* abstract value.
> > I get you want L3 MTU of 10 * 1500, but this value is not actually meaningful.
> >
> > IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
> > do not result in a trivial multiplication of the standard 1500 byte
> > ethernet L3 MTU.
> > Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
> > frames depending on which type of aggregation you do.
> > (and for tcp it even depends on the number and size of tcp options,
> > though it is often assumed that those take up 12 bytes, since that's the
> > normal for Linux-to-Linux tcp connections)
> >
> > For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu frames,
> > this means you have
> > N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> > payload (1500-12-20-40=1500-72=1428)
> > post aggregation:
> > 1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> > payload (N*1428)
> >
> > so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)
> >
> > That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
> > it's 40/60 if there's no tcp options (which I think happens when
> > talking to windows)
> > it's different still with ipv4 fragmentation... and again different
> > with ipv6 fragmentation...
> > etc.
> >
> > ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
> > Either way you don't get full frames.
> >
> > As such I'd recommend going with whatever is the largest mtu that can
> > be meaningfully made to fit in 16K with all the NCM header overhead.
> > That's likely closer to 15500-16000 (though I have *not* checked).
> >
> >> But my commit text didn't mention this
> >> properly which was a mistake on my behalf. But when I looked at the
> >> code, limiting the max segment size 15014 would force the practical
> >> max_mtu to not cross 15000 although theoretical max_mtu was set to:
> >> (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
> >>
> >> So my assumption of limiting it to 15000 was wrong. It must be limited
> >> to 15412 as mentioned in u_ether.c  This inturn means we must limit
> >> max_segment_size to:
> >> GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
> >> as mentioned in u_ether.c.
> >>
> >> I wanted to confirm that setting MAX_DATAGRAM_SIZE to
> >> GETHER_MAX_ETH_FRAME_LEN was correct.
> >>
> >> 2. I am not actually able to test with MTU beyond 15000. When my host
> >> device is a linux machine, the cdc_ncm.c limits max_segment_size to:
> >> CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */
> >
> > In practice you get 50% of the benefits of infinitely large mtu by
> > going from 1500 to ~2980.
> > you get 75% of the benefits by going to ~6K
> > you get 87.5% of the benefits by going to ~12K
> > the benefits of going even higher are smaller and smaller...
> >  > If the host side is limited to 8192, maybe we should match that here too?
>
> Hi Maciej,
>
>   Thanks for the detailed explanation. I agree with you on setting
> device side also to 8192 instead of what max_mtu is present in u_ether
> or practical max segment size possible.
>
> >
> > But the host side limitation of 8192 doesn't seem particularly sane either...
> > Maybe we should relax that instead?
> >
> I really didn't understand why it was set to 8192 in first place.
>
> > (especially since for things like tcp zero copy you want an mtu which
> > is slighly more then N * 4096,
> > ie. around 4.5KB, 8.5KB, 12.5KB or something like that)
> >
>
> I am not sure about host mode completely. If we want to increase though,
> just increasing the MAX_DATAGRAM_SIZE to some bigger value help ? (I
> don't know the entire code of cdc_ncm, so I might be wrong).
>
> Regards,
> Krishna,

Hmm, I'm not sure.  I know I've experimented with high mtu ncm in the past
(around 2.5 years ago).  I got it working between my Linux desktop (host)
and a Pixel 6 (device/gadget) with absolutely no problems.

I'm pretty sure I didn't change my desktop kernel, so I was probably
limited to 8192 there
(and I do more or less remember that).
Krishna Kurapati Oct. 14, 2023, 7:02 a.m. UTC | #10
On 10/14/2023 4:05 AM, Maciej Żenczykowski wrote:
>>>> The intent of posting the diff was two fold:
>>>>
>>>> 1. The question Greg asked regarding why the max segment size was
>>>> limited to 15014 was valid. When I thought about it, I actually wanted
>>>> to limit the max MTU to 15000, so the max segment size automatically
>>>> needs to be limited to 15014.
>>>
>>> Note that this is a *very* abstract value.
>>> I get you want L3 MTU of 10 * 1500, but this value is not actually meaningful.
>>>
>>> IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
>>> do not result in a trivial multiplication of the standard 1500 byte
>>> ethernet L3 MTU.
>>> Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
>>> frames depending on which type of aggregation you do.
>>> (and for tcp it even depends on the number and size of tcp options,
>>> though it is often assumed that those take up 12 bytes, since that's the
>>> normal for Linux-to-Linux tcp connections)
>>>
>>> For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu frames,
>>> this means you have
>>> N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
>>> payload (1500-12-20-40=1500-72=1428)
>>> post aggregation:
>>> 1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
>>> payload (N*1428)
>>>
>>> so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)
>>>
>>> That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
>>> it's 40/60 if there's no tcp options (which I think happens when
>>> talking to windows)
>>> it's different still with ipv4 fragmentation... and again different
>>> with ipv6 fragmentation...
>>> etc.
>>>
>>> ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
>>> Either way you don't get full frames.
>>>
>>> As such I'd recommend going with whatever is the largest mtu that can
>>> be meaningfully made to fit in 16K with all the NCM header overhead.
>>> That's likely closer to 15500-16000 (though I have *not* checked).
>>>
>>>> But my commit text didn't mention this
>>>> properly which was a mistake on my behalf. But when I looked at the
>>>> code, limiting the max segment size 15014 would force the practical
>>>> max_mtu to not cross 15000 although theoretical max_mtu was set to:
>>>> (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
>>>>
>>>> So my assumption of limiting it to 15000 was wrong. It must be limited
>>>> to 15412 as mentioned in u_ether.c  This inturn means we must limit
>>>> max_segment_size to:
>>>> GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
>>>> as mentioned in u_ether.c.
>>>>
>>>> I wanted to confirm that setting MAX_DATAGRAM_SIZE to
>>>> GETHER_MAX_ETH_FRAME_LEN was correct.
>>>>
>>>> 2. I am not actually able to test with MTU beyond 15000. When my host
>>>> device is a linux machine, the cdc_ncm.c limits max_segment_size to:
>>>> CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */
>>>
>>> In practice you get 50% of the benefits of infinitely large mtu by
>>> going from 1500 to ~2980.
>>> you get 75% of the benefits by going to ~6K
>>> you get 87.5% of the benefits by going to ~12K
>>> the benefits of going even higher are smaller and smaller...
>>>   > If the host side is limited to 8192, maybe we should match that here too?
>>
>> Hi Maciej,
>>
>>    Thanks for the detailed explanation. I agree with you on setting
>> device side also to 8192 instead of what max_mtu is present in u_ether
>> or practical max segment size possible.
>>
>>>
>>> But the host side limitation of 8192 doesn't seem particularly sane either...
>>> Maybe we should relax that instead?
>>>
>> I really didn't understand why it was set to 8192 in first place.
>>
>>> (especially since for things like tcp zero copy you want an mtu which
>>> is slighly more then N * 4096,
>>> ie. around 4.5KB, 8.5KB, 12.5KB or something like that)
>>>
>>
>> I am not sure about host mode completely. If we want to increase though,
>> just increasing the MAX_DATAGRAM_SIZE to some bigger value help ? (I
>> don't know the entire code of cdc_ncm, so I might be wrong).
>>
>> Regards,
>> Krishna,
> 
> Hmm, I'm not sure.  I know I've experimented with high mtu ncm in the past
> (around 2.5 years ago).  I got it working between my Linux desktop (host)
> and a Pixel 6 (device/gadget) with absolutely no problems.
> 
> I'm pretty sure I didn't change my desktop kernel, so I was probably
> limited to 8192 there
> (and I do more or less remember that).
>  From what I vaguely remember, it wasn't difficult (at all) to hit
> upwards of 7gbps for iperf tests.
> I don't remember how close to the theoretical USB 10gbps maximum of
> 9.7gbps I could get...
> [this was never the real bottleneck / issue, so I didn't ever dig
> particularly deep]
> 
> I'm pretty sure my gadget side changes were non-configurable...
> Probably just bumped one or two constants...
> 
Could you share what parameters you changed to get this high value of 
iperf throughput.

> I do *very* *vaguely* recall there being some funkiness though, where 8192 was
> *less* efficient than some slightly smaller value.
> 
> If I recall correctly the issue is that 8192 + ethernet overhead + NCM
> overhead only fits *once* into 16384, which leaves a lot of space
> wasted.
> While ~7.5 kb + overhead fits twice and is thus a fair bit better.
>Right, same goes for using 5K vs 5.5K MTU. If MTU is 5K, 3 packets can 
conveniently fit into an NTB but if its 5.5, at max only two (5.5k) 
packets can fit in (essentially filling ~11k of the 16384 bytes and 
wasting the rest)

And whether its Ipv4/Ipv6 like you mentioned on [1], the MTU is what NCM 
layer receives and we append the Ethernet header and add NCM headers and 
send it out after aggregation. Why can't we set the MAX_DATAGRAM_SIZE to 
~8050 or ~8100 ? The reason I say this value is, obviously setting it to 
8192 would not efficiently use the NTB buffer. We need to fill as much 
space in buffer as possible and assuming that each packet received on 
ncm layer is of MTU size set (not less that that), we can assume that 
even if only 2 packets are aggregated (minimum aggregation possible), we 
would be filling (2 * (8050 + ETH_HLEN) + (room for NCM headers)) would 
almost be close to 16384 ep max packet size. I already check 8050 MTU 
and it works. We can add a comment in code detailing the above 
explanation and why we chose to use 8050 or 8100 as MAX_DATAGRAM_SIZE.

Hope my reasoning of why we can chose 8.1K or 8.05K makes sense. Let me 
know your thoughts on this.

Regards,
Krishna,
Krishna Kurapati Oct. 14, 2023, 8:23 a.m. UTC | #11
On 10/14/2023 12:32 PM, Krishna Kurapati PSSNV wrote:
> 
> 
> On 10/14/2023 4:05 AM, Maciej Żenczykowski wrote:
>>>>> The intent of posting the diff was two fold:
>>>>>
>>>>> 1. The question Greg asked regarding why the max segment size was
>>>>> limited to 15014 was valid. When I thought about it, I actually wanted
>>>>> to limit the max MTU to 15000, so the max segment size automatically
>>>>> needs to be limited to 15014.
>>>>
>>>> Note that this is a *very* abstract value.
>>>> I get you want L3 MTU of 10 * 1500, but this value is not actually 
>>>> meaningful.
>>>>
>>>> IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
>>>> do not result in a trivial multiplication of the standard 1500 byte
>>>> ethernet L3 MTU.
>>>> Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
>>>> frames depending on which type of aggregation you do.
>>>> (and for tcp it even depends on the number and size of tcp options,
>>>> though it is often assumed that those take up 12 bytes, since that's 
>>>> the
>>>> normal for Linux-to-Linux tcp connections)
>>>>
>>>> For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu 
>>>> frames,
>>>> this means you have
>>>> N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
>>>> payload (1500-12-20-40=1500-72=1428)
>>>> post aggregation:
>>>> 1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
>>>> payload (N*1428)
>>>>
>>>> so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)
>>>>
>>>> That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
>>>> it's 40/60 if there's no tcp options (which I think happens when
>>>> talking to windows)
>>>> it's different still with ipv4 fragmentation... and again different
>>>> with ipv6 fragmentation...
>>>> etc.
>>>>
>>>> ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
>>>> Either way you don't get full frames.
>>>>
>>>> As such I'd recommend going with whatever is the largest mtu that can
>>>> be meaningfully made to fit in 16K with all the NCM header overhead.
>>>> That's likely closer to 15500-16000 (though I have *not* checked).
>>>>
>>>>> But my commit text didn't mention this
>>>>> properly which was a mistake on my behalf. But when I looked at the
>>>>> code, limiting the max segment size 15014 would force the practical
>>>>> max_mtu to not cross 15000 although theoretical max_mtu was set to:
>>>>> (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
>>>>>
>>>>> So my assumption of limiting it to 15000 was wrong. It must be limited
>>>>> to 15412 as mentioned in u_ether.c  This inturn means we must limit
>>>>> max_segment_size to:
>>>>> GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
>>>>> as mentioned in u_ether.c.
>>>>>
>>>>> I wanted to confirm that setting MAX_DATAGRAM_SIZE to
>>>>> GETHER_MAX_ETH_FRAME_LEN was correct.
>>>>>
>>>>> 2. I am not actually able to test with MTU beyond 15000. When my host
>>>>> device is a linux machine, the cdc_ncm.c limits max_segment_size to:
>>>>> CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */
>>>>
>>>> In practice you get 50% of the benefits of infinitely large mtu by
>>>> going from 1500 to ~2980.
>>>> you get 75% of the benefits by going to ~6K
>>>> you get 87.5% of the benefits by going to ~12K
>>>> the benefits of going even higher are smaller and smaller...
>>>>   > If the host side is limited to 8192, maybe we should match that 
>>>> here too?
>>>
>>> Hi Maciej,
>>>
>>>    Thanks for the detailed explanation. I agree with you on setting
>>> device side also to 8192 instead of what max_mtu is present in u_ether
>>> or practical max segment size possible.
>>>
>>>>
>>>> But the host side limitation of 8192 doesn't seem particularly sane 
>>>> either...
>>>> Maybe we should relax that instead?
>>>>
>>> I really didn't understand why it was set to 8192 in first place.
>>>
>>>> (especially since for things like tcp zero copy you want an mtu which
>>>> is slighly more then N * 4096,
>>>> ie. around 4.5KB, 8.5KB, 12.5KB or something like that)
>>>>
>>>
>>> I am not sure about host mode completely. If we want to increase though,
>>> just increasing the MAX_DATAGRAM_SIZE to some bigger value help ? (I
>>> don't know the entire code of cdc_ncm, so I might be wrong).
>>>
>>> Regards,
>>> Krishna,
>>
>> Hmm, I'm not sure.  I know I've experimented with high mtu ncm in the 
>> past
>> (around 2.5 years ago).  I got it working between my Linux desktop (host)
>> and a Pixel 6 (device/gadget) with absolutely no problems.
>>
>> I'm pretty sure I didn't change my desktop kernel, so I was probably
>> limited to 8192 there
>> (and I do more or less remember that).
>>  From what I vaguely remember, it wasn't difficult (at all) to hit
>> upwards of 7gbps for iperf tests.
>> I don't remember how close to the theoretical USB 10gbps maximum of
>> 9.7gbps I could get...
>> [this was never the real bottleneck / issue, so I didn't ever dig
>> particularly deep]
>>
>> I'm pretty sure my gadget side changes were non-configurable...
>> Probably just bumped one or two constants...
>>
> Could you share what parameters you changed to get this high value of 
> iperf throughput.
> 
>> I do *very* *vaguely* recall there being some funkiness though, where 
>> 8192 was
>> *less* efficient than some slightly smaller value.
>>
>> If I recall correctly the issue is that 8192 + ethernet overhead + NCM
>> overhead only fits *once* into 16384, which leaves a lot of space
>> wasted.
>> While ~7.5 kb + overhead fits twice and is thus a fair bit better.
> Right, same goes for using 5K vs 5.5K MTU. If MTU is 5K, 3 packets can 
> conveniently fit into an NTB but if its 5.5, at max only two (5.5k) 
> packets can fit in (essentially filling ~11k of the 16384 bytes and 
> wasting the rest)

Formatting gone wrong. So pasting the first paragraph again here:

"Right, same goes for using 5K vs 5.5K MTU. If MTU is 5K, 3 packets can
conveniently fit into an NTB but if its 5.5, at max only two (5.5k)
packets can fit in (essentially filling ~11k of the 16384 bytes and
wasting the rest)"

> 
> And whether its Ipv4/Ipv6 like you mentioned on [1], the MTU is what NCM 
> layer receives and we append the Ethernet header and add NCM headers and 
> send it out after aggregation. Why can't we set the MAX_DATAGRAM_SIZE to 
> ~8050 or ~8100 ? The reason I say this value is, obviously setting it to 
> 8192 would not efficiently use the NTB buffer. We need to fill as much 
> space in buffer as possible and assuming that each packet received on 
> ncm layer is of MTU size set (not less that that), we can assume that 
> even if only 2 packets are aggregated (minimum aggregation possible), we 
> would be filling (2 * (8050 + ETH_HLEN) + (room for NCM headers)) would 
> almost be close to 16384 ep max packet size. I already check 8050 MTU 
> and it works. We can add a comment in code detailing the above 
> explanation and why we chose to use 8050 or 8100 as MAX_DATAGRAM_SIZE.
> 
> Hope my reasoning of why we can chose 8.1K or 8.05K makes sense. Let me 
> know your thoughts on this.
> 

[1]: 
https://lore.kernel.org/all/CANP3RGd4G4dkMOyg6wSX29NYP2mp=LhMhmZpoG=rgoCz=bh1=w@mail.gmail.com/

> Regards,
> Krishna,
>
Maciej Żenczykowski Oct. 16, 2023, 1:19 a.m. UTC | #12
On Sat, Oct 14, 2023 at 1:24 AM Krishna Kurapati PSSNV
<quic_kriskura@quicinc.com> wrote:
>
>
>
> On 10/14/2023 12:32 PM, Krishna Kurapati PSSNV wrote:
> >
> >
> > On 10/14/2023 4:05 AM, Maciej Żenczykowski wrote:
> >>>>> The intent of posting the diff was two fold:
> >>>>>
> >>>>> 1. The question Greg asked regarding why the max segment size was
> >>>>> limited to 15014 was valid. When I thought about it, I actually wanted
> >>>>> to limit the max MTU to 15000, so the max segment size automatically
> >>>>> needs to be limited to 15014.
> >>>>
> >>>> Note that this is a *very* abstract value.
> >>>> I get you want L3 MTU of 10 * 1500, but this value is not actually
> >>>> meaningful.
> >>>>
> >>>> IPv4/IPv6 fragmentation and IPv4/IPv6 TCP segmentation
> >>>> do not result in a trivial multiplication of the standard 1500 byte
> >>>> ethernet L3 MTU.
> >>>> Indeed aggregating 2 1500 L3 mtu frames results in *different* sized
> >>>> frames depending on which type of aggregation you do.
> >>>> (and for tcp it even depends on the number and size of tcp options,
> >>>> though it is often assumed that those take up 12 bytes, since that's
> >>>> the
> >>>> normal for Linux-to-Linux tcp connections)
> >>>>
> >>>> For example if you aggregate N standard Linux ipv6/tcp L3 1500 mtu
> >>>> frames,
> >>>> this means you have
> >>>> N frames: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> >>>> payload (1500-12-20-40=1500-72=1428)
> >>>> post aggregation:
> >>>> 1 frame: ethernet (14) + ipv6 (40) + tcp (20) + tcp options (12) +
> >>>> payload (N*1428)
> >>>>
> >>>> so N * 1500 == N * (72 + 1428) --> 1 * (72 + N * 1428)
> >>>>
> >>>> That value of 72 is instead 52 for 'standard Linux ipv4/tcp),
> >>>> it's 40/60 if there's no tcp options (which I think happens when
> >>>> talking to windows)
> >>>> it's different still with ipv4 fragmentation... and again different
> >>>> with ipv6 fragmentation...
> >>>> etc.
> >>>>
> >>>> ie. 15000 L3 mtu is exactly as meaningless as 14000 L3 mtu.
> >>>> Either way you don't get full frames.
> >>>>
> >>>> As such I'd recommend going with whatever is the largest mtu that can
> >>>> be meaningfully made to fit in 16K with all the NCM header overhead.
> >>>> That's likely closer to 15500-16000 (though I have *not* checked).
> >>>>
> >>>>> But my commit text didn't mention this
> >>>>> properly which was a mistake on my behalf. But when I looked at the
> >>>>> code, limiting the max segment size 15014 would force the practical
> >>>>> max_mtu to not cross 15000 although theoretical max_mtu was set to:
> >>>>> (GETHER_MAX_MTU_SIZE - 15412) during registration of net device.
> >>>>>
> >>>>> So my assumption of limiting it to 15000 was wrong. It must be limited
> >>>>> to 15412 as mentioned in u_ether.c  This inturn means we must limit
> >>>>> max_segment_size to:
> >>>>> GETHER_MAX_ETH_FRAME_LEN (GETHER_MAX_MTU_SIZE + ETH_HLEN)
> >>>>> as mentioned in u_ether.c.
> >>>>>
> >>>>> I wanted to confirm that setting MAX_DATAGRAM_SIZE to
> >>>>> GETHER_MAX_ETH_FRAME_LEN was correct.
> >>>>>
> >>>>> 2. I am not actually able to test with MTU beyond 15000. When my host
> >>>>> device is a linux machine, the cdc_ncm.c limits max_segment_size to:
> >>>>> CDC_NCM_MAX_DATAGRAM_SIZE               8192    /* bytes */
> >>>>
> >>>> In practice you get 50% of the benefits of infinitely large mtu by
> >>>> going from 1500 to ~2980.
> >>>> you get 75% of the benefits by going to ~6K
> >>>> you get 87.5% of the benefits by going to ~12K
> >>>> the benefits of going even higher are smaller and smaller...
> >>>>   > If the host side is limited to 8192, maybe we should match that
> >>>> here too?
> >>>
> >>> Hi Maciej,
> >>>
> >>>    Thanks for the detailed explanation. I agree with you on setting
> >>> device side also to 8192 instead of what max_mtu is present in u_ether
> >>> or practical max segment size possible.
> >>>
> >>>>
> >>>> But the host side limitation of 8192 doesn't seem particularly sane
> >>>> either...
> >>>> Maybe we should relax that instead?
> >>>>
> >>> I really didn't understand why it was set to 8192 in first place.
> >>>
> >>>> (especially since for things like tcp zero copy you want an mtu which
> >>>> is slighly more then N * 4096,
> >>>> ie. around 4.5KB, 8.5KB, 12.5KB or something like that)
> >>>>
> >>>
> >>> I am not sure about host mode completely. If we want to increase though,
> >>> just increasing the MAX_DATAGRAM_SIZE to some bigger value help ? (I
> >>> don't know the entire code of cdc_ncm, so I might be wrong).
> >>>
> >>> Regards,
> >>> Krishna,
> >>
> >> Hmm, I'm not sure.  I know I've experimented with high mtu ncm in the
> >> past
> >> (around 2.5 years ago).  I got it working between my Linux desktop (host)
> >> and a Pixel 6 (device/gadget) with absolutely no problems.
> >>
> >> I'm pretty sure I didn't change my desktop kernel, so I was probably
> >> limited to 8192 there
> >> (and I do more or less remember that).
> >>  From what I vaguely remember, it wasn't difficult (at all) to hit
> >> upwards of 7gbps for iperf tests.
> >> I don't remember how close to the theoretical USB 10gbps maximum of
> >> 9.7gbps I could get...
> >> [this was never the real bottleneck / issue, so I didn't ever dig
> >> particularly deep]
> >>
> >> I'm pretty sure my gadget side changes were non-configurable...
> >> Probably just bumped one or two constants...
> >>
> > Could you share what parameters you changed to get this high value of
> > iperf throughput.

Eh, I really don't remember, but it wasn't anything earth shattering.
Krishna Kurapati Oct. 16, 2023, 3:48 a.m. UTC | #13
On 10/16/2023 6:49 AM, Maciej Żenczykowski wrote:

>>>>
>>>> Hmm, I'm not sure.  I know I've experimented with high mtu ncm in the
>>>> past
>>>> (around 2.5 years ago).  I got it working between my Linux desktop (host)
>>>> and a Pixel 6 (device/gadget) with absolutely no problems.
>>>>
>>>> I'm pretty sure I didn't change my desktop kernel, so I was probably
>>>> limited to 8192 there
>>>> (and I do more or less remember that).
>>>>   From what I vaguely remember, it wasn't difficult (at all) to hit
>>>> upwards of 7gbps for iperf tests.
>>>> I don't remember how close to the theoretical USB 10gbps maximum of
>>>> 9.7gbps I could get...
>>>> [this was never the real bottleneck / issue, so I didn't ever dig
>>>> particularly deep]
>>>>
>>>> I'm pretty sure my gadget side changes were non-configurable...
>>>> Probably just bumped one or two constants...
>>>>
>>> Could you share what parameters you changed to get this high value of
>>> iperf throughput.
> 
> Eh, I really don't remember, but it wasn't anything earth shattering.
>  From what I recall it was just a matter of bumping mtu, and tweaking
> irq pinning to stronger cores.
> Indeed I'm not even certain that the mtu was required to be over 5gbps.
> Though I may be confusing some things, as at least some of the testing was done
> with the kernel's built in packet generator.
> 
>>>
>>>> I do *very* *vaguely* recall there being some funkiness though, where
>>>> 8192 was
>>>> *less* efficient than some slightly smaller value.
>>>>
>>>> If I recall correctly the issue is that 8192 + ethernet overhead + NCM
>>>> overhead only fits *once* into 16384, which leaves a lot of space
>>>> wasted.
>>>> While ~7.5 kb + overhead fits twice and is thus a fair bit better.
>>> Right, same goes for using 5K vs 5.5K MTU. If MTU is 5K, 3 packets can
>>> conveniently fit into an NTB but if its 5.5, at max only two (5.5k)
>>> packets can fit in (essentially filling ~11k of the 16384 bytes and
>>> wasting the rest)
>>
>> Formatting gone wrong. So pasting the first paragraph again here:
>>
>> "Right, same goes for using 5K vs 5.5K MTU. If MTU is 5K, 3 packets can
>> conveniently fit into an NTB but if its 5.5, at max only two (5.5k)
>> packets can fit in (essentially filling ~11k of the 16384 bytes and
>> wasting the rest)"
>>
>>>
>>> And whether its Ipv4/Ipv6 like you mentioned on [1], the MTU is what NCM
>>> layer receives and we append the Ethernet header and add NCM headers and
>>> send it out after aggregation. Why can't we set the MAX_DATAGRAM_SIZE to
>>> ~8050 or ~8100 ? The reason I say this value is, obviously setting it to
>>> 8192 would not efficiently use the NTB buffer. We need to fill as much
>>> space in buffer as possible and assuming that each packet received on
>>> ncm layer is of MTU size set (not less that that), we can assume that
>>> even if only 2 packets are aggregated (minimum aggregation possible), we
>>> would be filling (2 * (8050 + ETH_HLEN) + (room for NCM headers)) would
>>> almost be close to 16384 ep max packet size. I already check 8050 MTU
>>> and it works. We can add a comment in code detailing the above
>>> explanation and why we chose to use 8050 or 8100 as MAX_DATAGRAM_SIZE.
>>>
>>> Hope my reasoning of why we can chose 8.1K or 8.05K makes sense. Let me
>>> know your thoughts on this.
> 
> Maybe just use an L3 mtu of 8000 then?  That's a nice round number...
> But I'm also fine with 8050 or 8100.. though 8100 seems 'rounder'.
> 
> I'm not sure what the actual overhead is... I guess we control the
> overhead in one direction, but not in the other, and there could be
> some slop, so we need to be a little generous?
> 
>>>
Hi Maciej,

   Sure. Let's go with 8000 to leave some space for headers. And would 
add the following paragraph as comment for readers to understand why 
this value was set:

"Although max mtu as dictated by u_ether is 15412 bytes, setting 
max_segment_size to 15426 would not be efficient. If user chooses 
segment size to be (> 8192), then we can't aggregate more than one 
buffer in each NTB (assuming each packet coming from network layer is > 
8192 bytes) as ep maxpacket limit is 16384. So let max_segment_size be 
limited to 8000 to allow atleast 2 packets to be aggregated reducing 
wastage of NTB buffer space"

Hope that would be fine.

Regards,
Krishna,
diff mbox series

Patch

diff --git a/drivers/usb/gadget/function/f_ncm.c b/drivers/usb/gadget/function/f_ncm.c
index feccf4c8cc4f..eab297b22200 100644
--- a/drivers/usb/gadget/function/f_ncm.c
+++ b/drivers/usb/gadget/function/f_ncm.c
@@ -103,6 +103,8 @@  static inline struct f_ncm *func_to_ncm(struct usb_function *f)
 /* Delay for the transmit to wait before sending an unfilled NTB frame. */
 #define TX_TIMEOUT_NSECS	300000
 
+#define MAX_DATAGRAM_SIZE	15014
+
 #define FORMATS_SUPPORTED	(USB_CDC_NCM_NTB16_SUPPORTED |	\
 				 USB_CDC_NCM_NTB32_SUPPORTED)
 
@@ -1408,6 +1410,7 @@  static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
 	ncm_opts = container_of(f->fi, struct f_ncm_opts, func_inst);
 
 	if (cdev->use_os_string) {
+		ncm_opts->net->mtu = (ncm_opts->max_segment_size - ETH_HLEN);
 		f->os_desc_table = kzalloc(sizeof(*f->os_desc_table),
 					   GFP_KERNEL);
 		if (!f->os_desc_table)
@@ -1469,6 +1472,8 @@  static int ncm_bind(struct usb_configuration *c, struct usb_function *f)
 
 	status = -ENODEV;
 
+	ecm_desc.wMaxSegmentSize = ncm_opts->max_segment_size;
+
 	/* allocate instance-specific endpoints */
 	ep = usb_ep_autoconfig(cdev->gadget, &fs_ncm_in_desc);
 	if (!ep)
@@ -1569,11 +1574,56 @@  USB_ETHERNET_CONFIGFS_ITEM_ATTR_QMULT(ncm);
 /* f_ncm_opts_ifname */
 USB_ETHERNET_CONFIGFS_ITEM_ATTR_IFNAME(ncm);
 
+static ssize_t ncm_opts_max_segment_size_show(struct config_item *item,
+					      char *page)
+{
+	struct f_ncm_opts *opts = to_f_ncm_opts(item);
+	u32 segment_size;
+
+	mutex_lock(&opts->lock);
+	segment_size = opts->max_segment_size;
+	mutex_unlock(&opts->lock);
+
+	return sprintf(page, "%u\n", segment_size);
+}
+
+static ssize_t ncm_opts_max_segment_size_store(struct config_item *item,
+					       const char *page, size_t len)
+{
+	struct f_ncm_opts *opts = to_f_ncm_opts(item);
+	int ret;
+	u32 segment_size;
+
+	mutex_lock(&opts->lock);
+	if (opts->refcnt) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	ret = kstrtou32(page, 0, &segment_size);
+	if (ret)
+		goto out;
+
+	if (segment_size > MAX_DATAGRAM_SIZE) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	opts->max_segment_size = segment_size;
+	ret = len;
+out:
+	mutex_unlock(&opts->lock);
+	return ret;
+}
+
+CONFIGFS_ATTR(ncm_opts_, max_segment_size);
+
 static struct configfs_attribute *ncm_attrs[] = {
 	&ncm_opts_attr_dev_addr,
 	&ncm_opts_attr_host_addr,
 	&ncm_opts_attr_qmult,
 	&ncm_opts_attr_ifname,
+	&ncm_opts_attr_max_segment_size,
 	NULL,
 };
 
@@ -1616,6 +1666,7 @@  static struct usb_function_instance *ncm_alloc_inst(void)
 		kfree(opts);
 		return ERR_CAST(net);
 	}
+	opts->max_segment_size = cpu_to_le16(ETH_FRAME_LEN);
 	INIT_LIST_HEAD(&opts->ncm_os_desc.ext_prop);
 
 	descs[0] = &opts->ncm_os_desc;
diff --git a/drivers/usb/gadget/function/u_ncm.h b/drivers/usb/gadget/function/u_ncm.h
index 5408854d8407..d3403cf13f17 100644
--- a/drivers/usb/gadget/function/u_ncm.h
+++ b/drivers/usb/gadget/function/u_ncm.h
@@ -31,6 +31,8 @@  struct f_ncm_opts {
 	 */
 	struct mutex			lock;
 	int				refcnt;
+
+	u32				max_segment_size;
 };
 
 #endif /* U_NCM_H */