[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <cdfc6f98-1aa0-4cb5-bd7d-93256552c39b@icloud.com>
Date: Tue, 27 Aug 2024 19:53:55 +0800
From: Zijun Hu <zijun_hu@...oud.com>
To: Jonathan Cameron <Jonathan.Cameron@...wei.com>
Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
"Rafael J. Wysocki" <rafael@...nel.org>, Davidlohr Bueso
<dave@...olabs.net>, Dave Jiang <dave.jiang@...el.com>,
Alison Schofield <alison.schofield@...el.com>,
Vishal Verma <vishal.l.verma@...el.com>, Ira Weiny <ira.weiny@...el.com>,
Dan Williams <dan.j.williams@...el.com>,
Takashi Sakamoto <o-takashi@...amocchi.jp>, Timur Tabi <timur@...nel.org>,
"David S. Miller" <davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
linux-kernel@...r.kernel.org, linux-cxl@...r.kernel.org,
netdev@...r.kernel.org, Zijun Hu <quic_zijuhu@...cinc.com>
Subject: Re: [PATCH v3 2/3] cxl/region: Find free cxl decoder by
device_for_each_child()
On 2024/8/27 19:30, Jonathan Cameron wrote:
> On Sat, 24 Aug 2024 17:07:44 +0800
> Zijun Hu <zijun_hu@...oud.com> wrote:
>
>> From: Zijun Hu <quic_zijuhu@...cinc.com>
>>
>> To prepare for constifying the following old driver core API:
>>
>> struct device *device_find_child(struct device *dev, void *data,
>> int (*match)(struct device *dev, void *data));
>> to new:
>> struct device *device_find_child(struct device *dev, const void *data,
>> int (*match)(struct device *dev, const void *data));
>>
>> The new API does not allow its match function (*match)() to modify
>> caller's match data @*data, but match_free_decoder() as the old API's
>> match function indeed modifies relevant match data, so it is not suitable
>> for the new API any more, solved by using device_for_each_child() to
>> implement relevant finding free cxl decoder function.
>>
>> Suggested-by: Ira Weiny <ira.weiny@...el.com>
>> Signed-off-by: Zijun Hu <quic_zijuhu@...cinc.com>
> This seems to functionally do the same as before.
>
yes, this change have the same logic as previous existing logic.
> I'm not sure I like the original code though so a comment inline.
>
>> ---
>> drivers/cxl/core/region.c | 30 ++++++++++++++++++++++++------
>> 1 file changed, 24 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
>> index 21ad5f242875..c2068e90bf2f 100644
>> --- a/drivers/cxl/core/region.c
>> +++ b/drivers/cxl/core/region.c
>> @@ -794,10 +794,15 @@ static size_t show_targetN(struct cxl_region *cxlr, char *buf, int pos)
>> return rc;
>> }
>>
>> +struct cxld_match_data {
>> + int id;
>> + struct device *target_device;
>> +};
>> +
>> static int match_free_decoder(struct device *dev, void *data)
>> {
>> + struct cxld_match_data *match_data = data;
>> struct cxl_decoder *cxld;
>> - int *id = data;
>>
>> if (!is_switch_decoder(dev))
>> return 0;
>> @@ -805,17 +810,31 @@ static int match_free_decoder(struct device *dev, void *data)
>> cxld = to_cxl_decoder(dev);
>>
>> /* enforce ordered allocation */
>> - if (cxld->id != *id)
>> + if (cxld->id != match_data->id)
>
> Why do we carry on in this case?
> Conditions are:
> 1. Start match_data->id == 0
> 2. First pass cxld->id == 0 (all good) or
> cxld->id == 1 say (and we skip until we match
> on cxld->id == 0 (perhaps on the second child if they are
> ordered (1, 0, 2) etc.
>
> If we skipped and then matched on second child but it was
> already in use (so region set), we will increment match_data->id to 1
> but never find that as it was the one we skipped.
>
> So this can only work if the children are ordered.
> So if that's the case and the line above is just a sanity check
> on that, it should be noisier (so an error print) and might
> as well fail as if it doesn't match all bets are off.
>
it seems Ira Weiny also has some concerns related to previous existing
logic as following:
https://lore.kernel.org/all/66c4a136d9764_2ddc2429435@iweiny-mobl.notmuch/
"Also for those working on CXL I'm questioning the use of ID here and
the dependence on the id's being added to the parent in order. Is that
a guarantee?"
perhaps, create a new dedicated thread to discuss original design.
> Jonathan
>
>> return 0;
>>
>> - if (!cxld->region)
>> + if (!cxld->region) {
>> + match_data->target_device = get_device(dev);
>> return 1;
>> + }
>>
>> - (*id)++;
>> + match_data->id++;
>>
>> return 0;
>> }
>>
>> +/* NOTE: need to drop the reference with put_device() after use. */
>> +static struct device *find_free_decoder(struct device *parent)
>> +{
>> + struct cxld_match_data match_data = {
>> + .id = 0,
>> + .target_device = NULL,
>> + };
>> +
>> + device_for_each_child(parent, &match_data, match_free_decoder);
>> + return match_data.target_device;
>> +}
>> +
>> static int match_auto_decoder(struct device *dev, void *data)
>> {
>> struct cxl_region_params *p = data;
>> @@ -840,7 +859,6 @@ cxl_region_find_decoder(struct cxl_port *port,
>> struct cxl_region *cxlr)
>> {
>> struct device *dev;
>> - int id = 0;
>>
>> if (port == cxled_to_port(cxled))
>> return &cxled->cxld;
>> @@ -849,7 +867,7 @@ cxl_region_find_decoder(struct cxl_port *port,
>> dev = device_find_child(&port->dev, &cxlr->params,
>> match_auto_decoder);
>> else
>> - dev = device_find_child(&port->dev, &id, match_free_decoder);
>> + dev = find_free_decoder(&port->dev);
>> if (!dev)
>> return NULL;
>> /*
>>
>
Powered by blists - more mailing lists