Claude review: fbdev: Wrap user-invoked calls to fb_set_var() in helper

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

**Bug: sh_mobile_lcdcfb loses "update all VCs" behavior**

The original `sh_mobile_fb_reconfig()` code was:

```c
if (fb_set_var(info, &var) < 0)
    return;

fbcon_update_vcs(info, true);  /* <-- always updates ALL VCs */
```

After this patch, it calls `fb_set_var_from_user(info, &var)`, which internally does:

```c
fbcon_update_vcs(info, var->activate & FB_ACTIVATE_ALL);
```

Since `var.activate = FB_ACTIVATE_NOW` (value 0) in `sh_mobile_fb_reconfig`, the expression `0 & 64` evaluates to `0` (false). This changes the behavior from updating **all** VCs to only the **current** VC.

When a display is re-plugged with a different mode, all consoles mapped to that framebuffer should be updated - the original `true` was intentional. The fix should be:

```c
var.activate = FB_ACTIVATE_NOW | FB_ACTIVATE_ALL;
```

in `sh_mobile_fb_reconfig()` before calling `fb_set_var_from_user`.

**ps3fb now gains fbcon_modechange_possible check** - This is a new pre-condition that wasn't present before. The cover letter says this is intentional ("fix inconsistencies"), which is reasonable, but worth noting as a behavioral change: if fbcon rejects the mode, ps3fb's ioctl will now return an error where it previously would have succeeded.

**Export level**: `EXPORT_SYMBOL(fb_set_var_from_user)` is consistent with `EXPORT_SYMBOL(fb_set_var)`. Correct, since ps3fb and sh_mobile_lcdcfb can be built as modules.

**Declaration placement**: `fb_set_var_from_user` is declared in `include/linux/fb.h` (public header), which is correct since it's called from drivers outside fbdev core.

The rest of the patch is mechanical and correct.

---
Generated by Claude Code Patch Reviewer