Hi, Thank you for sending the patch over! By the way, this is v2? If so, then it would be "[PATCH v2]", but not a big deal. A few nitpicks. > In function rockchip_pcie_host_init_port(), it defines a timeout > value of 500ms to wait for pcie training. However, it is not enough > for samsung PM953 SSD drive and realtek RTL8111F network adapter, > which leads to the following errors: It would be "PCIe" not "pcie", and "Samsung" and "Realtek" so that these are properly capitalised. > [ 0.879663] rockchip-pcie f8000000.pcie: PCIe link training gen1 timeout! > [ 0.880284] rockchip-pcie f8000000.pcie: deferred probe failed > [ 0.880932] rockchip-pcie: probe of f8000000.pcie failed with error -110 Normally, the time stamp is better removed as it's not very useful. > The pcie spec only defines the min time of training, not the max > one. So set a proper timeout value is important. Change the value > to 1000ms will fix this bug. [...] Again, it would be "PCIe" here. Also, since you are mentioning the PCI specification, would you be able to provide either a quote or a location (standard revision, chapter, page, etc.) from the PCI SIG? Additionally, if this change fixes a previous commit, then it might be prudent to include the "Fixes:" field referring to it, as it's customary to do so - a link to a relevant Bugzilla (https://bugzilla.kernel.org/) would also be fine, if there is anything. Other than that, Reviewed-by: Krzysztof Wilczyński <kw@xxxxxxxxx> Krzysztof
